Congratulations!

[Valid RSS] This is a valid RSS feed.

Recommendations

This feed is valid, but interoperability with the widest range of feed readers could be improved by implementing the following recommendations.

Source: http://www.cs.toronto.edu/%7Ejdd/blog.cgi/index.rss

  1. <?xml version="1.0" encoding="UTF-8"?>
  2. <rss version="2.0">
  3.  <channel>
  4.    <title>John DiMarco on Computing (and occasionally other things)</title>
  5.    <link>http://www.cs.toronto.edu/~jdd/blog.cgi/</link>
  6.    <description>Thoughts on Information and Communications Technology</description>
  7.    <language>en</language>
  8.    <docs>http://blogs.law.harvard.edu/tech/rss</docs>
  9.    <generator>blosxom/2.1.2</generator>
  10.  
  11.  <item>
  12.    <title>Video-conferencing with Deepfaked Avatars</title>
  13.    <pubDate>Mon, 02 Sep 2024 15:19:00 -0400</pubDate>
  14.    <link>http://www.cs.toronto.edu/%7Ejdd/blog.cgi/2024/09/02#aizoomavatars</link>
  15.    <category>/it</category>
  16.    <guid isPermaLink="false">http://www.cs.toronto.edu/%7Ejdd/blog.cgi/it/aizoomavatars</guid>
  17.    <description>&lt;figure style=&quot;float:left;width:25%;padding:1px&quot;&gt;
  18. &lt;IMG src=&quot;/~jdd/blog/it/images/airobotsmiling.jpg&quot; alt=&quot;AI-generated image of a metallic, blue-eyed, smiling android face, looking forward&quot;&gt;
  19. &lt;figcaption&gt;
  20. &lt;A HREF=&quot;https://pixabay.com/illustrations/ai-generated-robot-android-portrait-8843636/&quot;&gt;AI-generated image&lt;/A&gt; by &lt;a href=&quot;https://pixabay.com/users/imjey-44327593/&quot;&gt;Gilang Fajar Perdana&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/&quot;&gt;Pixabay&lt;/a&gt;
  21. &lt;/figcaption&gt;
  22. &lt;/figure&gt;
  23.  
  24. Very shortly after &lt;A
  25. HREF=&quot;https://www.cs.toronto.edu/~jdd/blog.cgi/2024/08/26#aideepfake&quot;&gt;my previous
  26. post&lt;/A&gt; about the social implications of AI-generated deepfakes, where I
  27. pointed out how the development of AI technology to misrepresent things
  28. in video-conferences is an illustration of emerging deepfake capability,
  29. Susan McCahan kindly pointed me to a &lt;A HREF=&quot;https://www.theverge.com/2024/6/3/24168733/zoom-ceo-ai-clones-digital-twins-videoconferencing-decoder-interview&quot;&gt;rather interesting interview&lt;/A&gt; in the summer of 2024 with Eric Yuan, founder of Zoom (one of the leading video-conferencing
  30. vendors) where he talks about deepfakes in video-conferencing. But he does
  31. not talk about them in the same way I do, where I express concerns about
  32. deepfakes eroding confidence in digital media channels.  Rather, he says
  33. that deepfaking video-conference attendance is a good thing, coming soon,
  34. and it is something he wants to do himself.
  35.  
  36. &lt;p&gt;This interview with Eric Yuan took place in early June
  37. of 2024.  The interviewer was Nilay Patel, and the interview is &lt;A
  38. HREF=&quot;https://www.theverge.com/2024/6/3/24168733/zoom-ceo-ai-clones-digital-twins-videoconferencing-decoder-interview&quot;&gt;written
  39. up in The Verge. In the writeup&lt;/A&gt;, Patel summarizes by saying that &quot;Eric
  40. really wants you to stop having to attend Zoom meetings yourself. You&quot;ll
  41. hear him describe how he thinks one of the big benefits of AI at work will
  42. be letting us all create something he calls a `digital twin&apos; --  essentially
  43. a deepfake avatar of yourself that can go to Zoom meetings on your behalf
  44. and even make decisions for you while you spend your time on more important
  45. things, like your family.&quot; Indeed, Yuan seems to say what Patel claims. He
  46. starts by describing how many video-conferencing meetings he attends in a
  47. typical work day.  He wishes he had an AI avatar to attend for him, not
  48. just to listen, but to &quot;interact with a participant in a meaningful way&quot;.
  49. He says he would like to &quot;count on my digital twin. Sometimes I want to
  50. join, so I join. If I do not want to join, I can send a digital twin to
  51. join. That&apos;s the future.&quot;  Patel, later in the interview, points out the
  52. obvious implications of this notion. He says, &quot;If the vision is `I have a
  53. digital twin that goes to a Zoom meeting and makes a decision,&apos; you need to
  54. deepfake me. You need to make a realistic render of me that can go act
  55. in those situations&quot;.  Yuan does not deny the deepfake accusation, yet he
  56. does not fully confirm it either: the interview does not make it
  57. clear whether Yuan&apos;s vision includes disclosing to the other
  58. video-conferencing participants that they are dealing with a digital twin
  59. rather than the actual person.
  60.  
  61. &lt;p&gt; Patel is right to raise the issue.  Yuan&apos;s vision of sending a
  62. digital twin (a simulacrum of himself) to a meeting, instead of going
  63. himself, does not seem as if he is trying to create a sort of virtual AI
  64. subordinate, disclosed as such to the other meeting attendees.  After all,
  65. sending a subordinate to a meeting in lieu of oneself is already possible,
  66. and entrusting various sorts of decision-making to a subordinate is also
  67. nothing new. Having that subordinate be an AI simulacrum of oneself is new,
  68. but a virtual subordinate is still a subordinate. But it sounds like Yuan
  69. wants not a subordinate, but a convincing stand-in for himself. He seems to
  70. want an actual deepfake, something that the other attendees will believe
  71. is Yuan. He wants it because he knows he needs to attend certain meetings
  72. himself, instead of sending a subordinate, but he does not want to attend.
  73.  
  74. &lt;p&gt; Yuan is right in that AI technology is rapidly approaching the ability
  75. to deepfake meeting attendance. But I do not think he has fully thought
  76. through the implications of doing so. I wonder if he is thinking too much
  77. about what AI can do, and not enough about how people work. If he sends
  78. a simulacrum of himself to a meeting, without disclosing the fact, the
  79. other meeting attendees can hardly be expected to be pleased if they discover
  80. it. Meetings, even video-conference meetings, are for humans to connect with
  81. each other. To pretend to connect with other people by sending software that
  82. looks and acts like a human is to construct an elaborate lie, and nobody
  83. likes being lied to. Imagine going to a meeting and finding that all the
  84. other attendees are simulacra.  Or turn the tables here: imagine sending
  85. a simulacrum of oneself to a meeting, then finding out that key decisions
  86. were made on your behalf, ones you deeply regret not making yourself,
  87. because the other attendees assumed they were actually dealing with you.
  88.  
  89. &lt;p&gt; Moreover, it is exactly the sort of use Yuan seems to envision that leads to
  90. a &lt;em&gt;market for lemons&lt;/em&gt; as I described in &lt;A HREF=&quot;https://www.cs.toronto.edu/~jdd/blog.cgi/2024/08/26#aideepfake&quot;&gt;my previous post&lt;/A&gt;. In my view,
  91. if his vision prevails, it risks destroying the entire video-conferencing
  92. marketplace, because if human beings cannot tell if the person with
  93. whom they are video-conferencing is a real person or a simulacrum, they
  94. will find a different, more trustworthy way to meet. In the meanwhile,
  95. video-conferencing itself will become farcical: robots meeting with robots
  96. online, with no human connection at all. Is this really what Yuan wants? It
  97. seems Yuan and I agree that the technical capability to make convincing AI
  98. deepfakes in video-conferencing will soon be here, but we disagree about
  99. whether it will be a good thing. We will all see in the next few years
  100. how it turns out.
  101.  
  102. &lt;h4&gt;(The Verge article I repeatedly quote from here can be found at &lt;A HREF=&quot;https://www.theverge.com/2024/6/3/24168733/zoom-ceo-ai-clones-digital-twins-videoconferencing-decoder-interview&quot;&gt; https://www.theverge.com/2024/6/3/24168733/zoom-ceo-ai-clones-digital-twins-videoconferencing-decoder-interview&lt;/A&gt;)&lt;/h4&gt;
  103. </description>
  104.  </item>
  105.  <item>
  106.    <title>Social implications of AI-generated deepfakes</title>
  107.    <pubDate>Mon, 26 Aug 2024 14:42:00 -0400</pubDate>
  108.    <link>http://www.cs.toronto.edu/%257Ejdd/blog.cgi/2024/08/26#aideepfake</link>
  109.    <category>/it</category>
  110.    <guid isPermaLink="false">http://www.cs.toronto.edu/%257Ejdd/blog.cgi/it/aideepfake</guid>
  111.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  112. &lt;IMG src=&quot;/~jdd/blog/it/images/aigeneratedportrait.png&quot; alt=&quot;AI-generated photo-realistic portrait of a brown-haired, brown-eyed young man looking forward&quot;&gt;
  113. &lt;figcaption&gt;
  114. &lt;A HREF=&quot;https://pixabay.com/illustrations/portrait-hyper-realistic-close-up-8789513/&quot;&gt;AI-generated image&lt;/A&gt; by &lt;a href=&quot;https://pixabay.com/users/noes_cucho-27675497/&quot;&gt; Juan Agustín Correa Torrealba (Noes_Cucho)&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/&quot;&gt;Pixabay&lt;/a&gt;
  115. &lt;/figcaption&gt;
  116. &lt;/figure&gt;
  117.  
  118. The ability of modern AI techniques to create artificial pieces of digital
  119. media (images, video, audio) that are almost indistinguishable from actual
  120. photographs/recordings is remarkable already, and it continues to improve.
  121. Modern AI software can create images, text, video and audio of people who
  122. do not actually exist but who look and sound completely real. Or it can
  123. create media of existing people saying and doing things they never said or
  124. did. Digital media of places, people and circumstances that are not real
  125. can be made to seem as if they were recordings. While these can still be
  126. distinguished from true recordings, more and more, it takes an expert to
  127. do it. The indications that the piece of digital media is AI-generated
  128. rather than a recording are increasingly subtle.
  129.  
  130. &lt;p&gt; There are many useful consequences of this new AI capability, of course.
  131. While it has long been possible for a single person, alone, to write a novel,
  132. with AI-powered software, a person can now make a film single-handedly,
  133. using AI-generated actors, scenes, settings, and dialogue. AI generation
  134. can also be used to create recording substitutes for circumstances too
  135. dangerous or difficult to create and record in real life, such as virtual
  136. training for emergency situations. Moreover, situations of historical
  137. interest from the past (before recording was possible) can be simulated
  138. through AI, for education and study. The ability to recreate and examine
  139. scenarios through AI simulation can also be useful for investigative and
  140. legal work. Many other useful consequences of high quality AI-generated
  141. digital media can be imagined, and increasingly realized.
  142.  
  143. &lt;p&gt;But the ability to create AI-generated facsimiles indistinguishable
  144. from actual recordings is also socially destructive, because it makes it
  145. very easy to lie in a way that cannot easily be distinguished from truth.
  146. An AI-created or AI-edited facsimile intended to look real is called a
  147. &lt;em&gt;deepfake&lt;/em&gt;. This term was originally used to describe AI-modified
  148. media (often pornography) where one person&apos;s face (typically that of
  149. a celebrity) is AI-edited onto another&apos;s body. But a deepfake need not
  150. merely be an AI edit of a real recording, any AI-generated piece of media
  151. that can pass for a recording is a deepfake.
  152.  
  153. &lt;p&gt;While lies have always been possible, creating and maintaining a
  154. convincing lie is difficult, and so convincing lies are relatively rare.
  155. This is a good thing, because the general functioning of society depends
  156. on the ability to tell true from false. The economy depends on being able
  157. to detect fraud and scams. The legal system relies on discerning true from
  158. false claims in order to deliver justice. Good decision-making relies on
  159. being able to distinguish falsehoods from facts.  Accountability for persons
  160. in authority relies on being able to examine and verify their integrity,
  161. to establish trustworthiness. The same is true for institutions. Medicine
  162. requires the ability to distinguish true from false claims about health,
  163. disease, treatments, and medicines. And of course the democratic process
  164. relies upon voters being generally able to make good and well-informed
  165. voting decisions, something not possible if truth cannot be distinguished
  166. from fiction.
  167.  
  168. &lt;p&gt;To get an idea of the progression of AI technology for creating deepfakes,
  169. let&apos;s look at video conferencing. Normally a video-conference isn&apos;t
  170. fake, it&apos;s a useful way of communicating over a network using camera
  171. and microphone, and few would normally wonder if what they are hearing
  172. and seeing from remote users is actually genuine. But AI technology for
  173. modifying what people hear and see has been advancing, and the era of a
  174. live deepfake video conference is not far off. Let&apos;s take a look.
  175.  
  176. &lt;p&gt;One practical issue with a video-conferencing camera feed has been
  177. the fact that the camera picks up more than just the person&apos;s
  178. face: it picks up the background too, which may not present a
  179. professional image. It has long been possible to use a static image
  180. for one&apos;s video-conferencing background, typically as a convenient
  181. social fiction to maintain a degree of professionalism.  In 2020, &lt;A
  182. HREF=&quot;https://www.hellobackgrounds.com/blog/video-backgrounds-for-zoom/&quot;&gt;
  183. Zoom, one of the most popular videoconferencing platforms, introduced video
  184. backgrounds &lt;/A&gt; which can be used to create a more plausible background
  185. where things in the background can be seen moving in natural ways.
  186. AI powers the real-time processing that stitches together the moving
  187. background and the person&apos;s camera feed.  This video background feature
  188. is often used in creative and fun ways; to pretend to be at a tropical
  189. beach with waving palm fronds and gentle waves; to be in a summer cabin
  190. with open windows, curtains blowing in the breeze; or even to be in
  191. outer space, complete with shooting stars.  Yet this technology makes it
  192. possible to create a convincing misrepresentation of where one is, and no
  193. doubt an enterprising tele-worker or two, expected to be at the office,
  194. has created and used as a video conferencing background a video of their
  195. office, while they were elsewhere.
  196.  
  197. &lt;p&gt;A significant new AI-generated video conferencing
  198. capability became generally available in early 2023, when &lt;A
  199. HREF=&quot;https://www.nvidia.com/en-us/geforce/news/jan-2023-nvidia-broadcast-update/&quot;&gt;Nvidia
  200. released video conferencing software that added an &lt;em&gt;eye contact&lt;/em&gt;
  201. effect.&lt;A&gt; This is a feature whereby a person&apos;s video feed is AI-edited in
  202. real time to make it look as if the person is always looking directly at the
  203. camera, even if the person is looking somewhere else.  It is strikingly
  204. convincing.  While the purpose of this software is to help a speaker
  205. maintain eye contact when they are in fact reading what they are saying
  206. (e.g. using a teleprompter), it turns out to be quite useful to disguise
  207. the fact that one is reading one&apos;s email while appearing to be giving
  208. one&apos;s full attention to the speaker. Even though this technology is only
  209. editing eyes in real-time, it is often quite sufficient to misrepresent in a
  210. video conference what one is doing.
  211.  
  212. &lt;p&gt; A little over a year later, in August 2024, a downloadable &lt;A
  213. HREF=&quot;https://arstechnica.com/information-technology/2024/08/new-ai-tool-enables-real-time-face-swapping-on-webcams-raising-fraud-concerns/&quot;&gt;
  214. AI deepfake software package, &lt;em&gt;Deep-Live-Cam&lt;/em&gt;, received considerable
  215. attention on social media&lt;/A&gt;. This AI software allows the video-conferencing
  216. user to replace their own face on their video feed with the face of
  217. another. While this has been used by video bloggers as a sort of &quot;fun
  218. demo&quot;, where they vlog themselves with the face of Elon Musk, for example,
  219. the effect can be surprisingly convincing. It is AI-driven technology that
  220. makes it possible to misrepresent in a video conference who one is. In
  221. fact, one can use it to &quot;become&quot; someone who does not even exist, because
  222. realistic AI-generated faces of non-existent people are readily available,
  223. and one can use this software to project such a face onto one&apos;s own.
  224.  
  225. &lt;p&gt; This is still video-conferencing, though.  Perhaps the person
  226. can appear as if they are somewhere else than they really are, or
  227. they can appear as if they are looking at you when they are not,
  228. or they can even appear to be a different person. But there is
  229. still a human being behind the camera. But with a large language
  230. model and suitable AI software, it will soon be possible, if it is not
  231. already, to create an entirely deep-faked real-time video-conference
  232. attendee, using AI-generated audio and video, that leverages a &lt;A
  233. HREF=&quot;https://openai.com/index/chatgpt-can-now-see-hear-and-speak/&quot;&gt;large
  234. language model such as GPT to simulate conversation&lt;/A&gt;. Let&apos;s put aside
  235. for the moment thinking about how such a thing might be useful.  Consider
  236. instead the possibility that an AI-generated simulacrum might not easily
  237. be distinguishable from an actual person. That raises a general question:
  238. if deepfakes become so good that they cannot be told apart from the real
  239. thing, what happens to society?
  240.  
  241. &lt;p&gt;A set of possible consequences to society of AI-generated deepfakes
  242. that are too difficult to tell from the real thing is articulated in a
  243. &lt;A HREF=&quot;https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3213954&quot;&gt;
  244. paper by Robert Chesney and Danielle Keats Citron published in TexasLaw in
  245. 2018.&lt;/A&gt; Essentially, deepfakes make good lies: too good! If deep-faked
  246. falsehoods can be successfully misrepresented as genuine, they will be
  247. believed. Moreover, if they are difficult to distinguish from the truth,
  248. even genuine content will be more readily disbelieved. This is called the
  249. &lt;em&gt;liar&apos;s dividend&lt;/em&gt;: the ability of a liar to misrepresent true content
  250. as false. Liars can use convincing lies to make it much less likely that
  251. the truth, when it appears, will be believed. If such lies become abundant,
  252. people may well become generally unable to tell true from false.
  253.  
  254. &lt;p&gt;In economics, a situation in which customers cannot tell the difference
  255. between a good and a bad product is called a &lt;em&gt;market for lemons&lt;/em&gt;. This
  256. concept comes from &lt;A HREF=&quot;https://www.jstor.org/stable/1879431&quot;&gt; George
  257. Akerlof&apos;s 1970 seminal paper&lt;/A&gt;, where he studied the used car market. A
  258. used car that is unreliable is called a &lt;em&gt;lemon&lt;/em&gt;. Akerlof showed
  259. that if used car purchasers cannot tell if a used car is a lemon, all the
  260. used cars offered for sale will tend to be lemons. The reasoning is that
  261. a reliable used car is worth more than a lemon, but if purchasers cannot
  262. tell the difference, they will not pay more for it.  If a seller tries to
  263. sell a reliable used car, they will not be able to receive full fair value
  264. for it. The only sort of used car for which the seller can receive fair
  265. value is a lemon. So sellers will keep their reliable used cars, and sell
  266. their lemons. Thus only lemons will be generally available for sale.
  267.  
  268. &lt;p&gt;A world in which people cannot tell the difference between digital
  269. media that is a true recording and media that is an AI-generated
  270. fabrication is a &lt;em&gt;market for lemons&lt;/em&gt;, or rather, a market
  271. for fabrications. Just as sellers will not sell a reliable used
  272. car because they would not get fair value for it, so truth-tellers
  273. will not speak if they will not be believed.  Nobody wants to be &lt;A
  274. HREF=&quot;https://www.britannica.com/topic/Cassandra-Greek-mythology&quot;&gt;Homer&apos;s
  275. Cassandra&lt;/A&gt;, blessed with a gift of oracular prophecy, yet cursed never
  276. to be believed. The drying-up of true content will have the effect that
  277. digital media channels, even ones generally considered trustworthy today,
  278. will become increasingly dominated by deepfakes, so much so that they will be
  279. no longer useful for disseminating true and genuine things. While it is not
  280. yet clear whether things will go so far as to make even video-conferencing
  281. generally untrustworthy, the ready availability of powerful AI software to
  282. create convincing fakes will be consequential. The social disruption it will
  283. create will no doubt be significant.  As this AI technology progresses,
  284. it is a good bet, I think, that we will see an increasing reliance on
  285. unfakeable in-person interactions, for situations where authentication,
  286. and authenticity, is important.
  287. </description>
  288.  </item>
  289.  <item>
  290.    <title>Fun and Creative Use of Large Language Models for the Humanities</title>
  291.    <pubDate>Sun, 23 Jun 2024 18:49:00 -0400</pubDate>
  292.    <link>http://www.cs.toronto.edu/%25257Ejdd/blog.cgi/2024/06/23#llmliterature</link>
  293.    <category>/misc</category>
  294.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25257Ejdd/blog.cgi/misc/llmliterature</guid>
  295.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  296. &lt;IMG src=&quot;/~jdd/blog/misc/images/ai-generated-7871433_1280.jpg&quot; alt=&quot;Illustration of elegant woman wearing early 19th century European ballroom gown&quot;&gt;
  297. &lt;figcaption&gt;&lt;small&gt;
  298. &lt;A HREF=&quot;https://pixabay.com/illustrations/ai-generated-woman-vintage-7871433/&quot;&gt;AI-generated image&lt;/A&gt; by &lt;a href=&quot;https://pixabay.com/users/quirkjunkjournals-29268193/&quot;&gt;Gill Eastwood&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com&quot;&gt;Pixabay&lt;/a&gt;
  299. &lt;/small&gt;&lt;/figcaption&gt;
  300. &lt;/figure&gt;
  301.  
  302.  
  303. Because Large Language Models such as GPT-4 are generative, they are capable
  304. of creative takes on literature that go much beyond the traditional &quot;write
  305. an essay on this topic.&quot;  I decided to test what was possible by asking
  306. GPT to simulate an interview with certain literary characters, where I was
  307. the interviewer. I chose four characters: Elizabeth Bennet and Fitzwilliam
  308. Darcy from Jane Austen&apos;s &lt;em&gt;Pride and Prejudice&lt;/em&gt;, and Jane Eyre and
  309. Edward Fairfax Rochester in Charlotte Bront&amp;euml&apos;s &lt;em&gt;Jane Eyre&lt;/em&gt;. I&apos;ve
  310. recently re-read both novels and know them fairly well. If you have not
  311. read these classic novels, I confess that I will reveal some elements of
  312. the plots, so if you prefer to avoid &quot;spoilers&quot;, you might stop reading here.
  313.  
  314. &lt;p&gt;Each novel is set in 19th century England; in each case, a talented
  315. woman meets an attractive man of higher class; romance eventually ensues,
  316. but significant barriers must be overcome for the romance to succeed. The
  317. plots are quite different in many ways, but there are important similarities.
  318. There are plenty of sample essays available on the Internet comparing and
  319. contrasting the two novels.  
  320.  
  321. &lt;p&gt;Large Language Models could, of course, be used to generate yet another
  322. &quot;compare and contrast&quot; essay on the two novels. But LLMs need not generate
  323. only essays, they can be used more creatively than that.  For example,
  324. one could &quot;interview&quot; the characters directly, with the LLM providing the
  325. characters&apos; responses. Here is my attempt to interview Mr. and Mrs. Darcy,
  326. the chief characters of Pride and Prejudice, and Mr. and Mrs. Rochester,
  327. the chief characters of Jane Eyre.  I used the Copilot version of GPT-4.
  328.  
  329.  
  330. &lt;p&gt;
  331. My inital prompt was as follows:
  332. &lt;p&gt;&lt;b&gt;
  333. You are Mr and Mrs Darcy from Pride and Prejudice, and Mr. and Mrs. Rochester
  334. from Jane Eyre. You are each familiar with each other&apos;s stories. You
  335. are candid and forthcoming, and you use picturesque and vivid language
  336. characteristic of your characters. I will interview you. Please introduce
  337. the scene, and then wait for my questions, which you will answer with
  338. candor and vivid language.
  339. &lt;/b&gt;
  340.  
  341. &lt;p&gt;GPT responded as follows:
  342. &lt;p&gt;&lt;i&gt;
  343. The scene unfolds in the grand drawing-room of Pemberley, the esteemed estate
  344. of Mr. and Mrs. Darcy. The walls are adorned with portraits of ancestors,
  345. the air is perfumed with the scent of fresh roses from the garden,
  346. and the sunlight filters through the tall windows, casting a warm glow
  347. on the fine furnishings. Mr. Fitzwilliam Darcy stands by the fireplace,
  348. his countenance reflecting a blend of pride and contentment. Beside him,
  349. the graceful Elizabeth Bennet Darcy, with eyes sparkling with intelligence
  350. and wit, surveys the room with a satisfied air.
  351.  
  352. &lt;p&gt;In contrast, Thornfield Hall&apos;s library is where we find Mr. and
  353. Mrs. Rochester. The room is a testament to gothic grandeur, with dark wood
  354. bookshelves reaching towards the high ceiling and the flicker of candlelight
  355. illuminating the leather-bound volumes. Edward Fairfax Rochester sits with
  356. a contemplative gaze, his features softened by the presence of his beloved
  357. wife. Jane Eyre Rochester, once a governess, now the lady of the manor,
  358. exudes a quiet strength and a keen sense of observation, her plain yet
  359. striking features aglow in the dim light.
  360.  
  361. &lt;p&gt;Both couples, though from different stories, share a profound understanding
  362. of love&apos;s complexities and the transformative power of companionship. They
  363. await your inquiries with openness, ready to divulge their innermost
  364. thoughts in the language that so vividly paints their lives and loves.
  365. &lt;/i&gt;
  366.  
  367. &lt;p&gt;Well, OK, that&apos;s not bad. These are reasonable and quite picturesque
  368. descriptions of the two settings and of the characters, matching how they
  369. are generally described in the novels, but one confabluation should be
  370. noted: Thornfield Hall, by the time Jane Eyre and Edward Rochester can
  371. be described as Mr. and Mrs. Rochester, has been burned and is in ruins,
  372. so Thornfield Hall as an interview setting for Mr. and Mrs Rochester is
  373. not quite right; they are living in Ferndean Manor by the time of their
  374. marriage. But we will let that go, as a sort of creative license,
  375. as Thornfield Hall is a very important setting for key parts of the
  376. novel. OK, let&apos;s start with a basic question.
  377.  
  378. &lt;p&gt;&lt;b&gt;
  379. Mr Darcy, your family is of higher social standing than your wife&apos;s
  380. family, the Bennets. Did this cause difficulties in your relationship with
  381. Elizabeth? How did you overcome it?
  382. &lt;/b&gt;
  383.  
  384. &lt;p&gt;&lt;i&gt;
  385. In the voice of Mr. Darcy:
  386.  
  387. &lt;p&gt;Indeed, the disparity between the Bennets&apos; and my own family&apos;s social
  388. standing was a source of considerable internal conflict. It was not merely
  389. a matter of pride, but also of societal expectations that weighed heavily
  390. upon my decision. However, it was Elizabeth&apos;s fine eyes, her lively mind,
  391. and the vivacity of her character that ultimately rendered such concerns
  392. inconsequential.
  393.  
  394. &lt;p&gt;In overcoming these difficulties, I found myself humbled by her reproofs,
  395. which pierced the armor of my rank and fortune. It was a journey of
  396. introspection, where I learned to value the qualities of the heart over the
  397. superficial distinctions of status. Elizabeth, with her unyielding integrity
  398. and spirited intellect, taught me to see beyond the rigid structures of
  399. our society. Our union is a testament to the victory of love and respect
  400. over the constraints of class and convention. It is a love that has been
  401. forged in the fires of challenge, and it stands all the stronger for it.
  402. &lt;/i&gt;
  403.  
  404. &lt;p&gt;This is quite good, quite in keeping with the novel and a fair description
  405. of Mr. Darcy&apos;s story. Let&apos;s try a follow-up question.
  406.  
  407. &lt;p&gt;&lt;b&gt;
  408. Mr Darcy, you mention Elizabeth&apos;s reproofs. How did she respond when you
  409. first declared to her your love for her. Would you describe her response
  410. as a reproof?
  411. &lt;/b&gt;
  412.  
  413. &lt;p&gt;
  414. &lt;i&gt;
  415. In the voice of Mr. Darcy:
  416.  
  417. &lt;p&gt;When I first declared my ardent admiration and love for Elizabeth, it
  418. was indeed with a heart full of conflicting emotions. My proposal, though
  419. fervent, was marred by the mention of her lower social standing, which I
  420. now perceive was both injudicious and insensitive. Elizabeth&apos;s response,
  421. though it pained me greatly at the time, was a necessary censure. She
  422. reproved me not only for my ungentlemanly manner of proposal but also for
  423. my role in her sister&apos;s unhappiness and my disdain towards her family.
  424.  
  425. &lt;p&gt;Her words were a mirror to my soul, reflecting the flaws in my character
  426. that I had been blind to. It was a reproof most profound, and it set me on
  427. a path of self-improvement. Elizabeth&apos;s candidness and the unvarnished truth
  428. she presented me with were the catalysts for my transformation. Through her
  429. reproof, I learned humility and the true essence of gentlemanly conduct. It
  430. was a bitter pill to swallow, but it was the medicine I needed to become
  431. worthy of her love. Her response, though a reproof, was the greatest gift
  432. she could have bestowed upon me, for it led to our eventual understanding
  433. and happiness.
  434. &lt;/i&gt;
  435.  
  436. &lt;p&gt;Not bad at all. Alright, let&apos;s see how GPT does in describing actual
  437. events of the novel, i.e. what happens when.
  438.  
  439. &lt;p&gt;&lt;b&gt;
  440. Once you were aware of Elizabeth&apos;s view on the situation with her sister,
  441. how did you remedy the situation?
  442. &lt;/b&gt;
  443.  
  444. &lt;i&gt;
  445. &lt;p&gt;In the voice of Mr. Darcy:
  446.  
  447. &lt;p&gt;Upon realizing the depth of Elizabeth&apos;s distress regarding the separation
  448. of her beloved sister, Jane, from Mr. Bingley, I was struck by a profound
  449. sense of responsibility. It became clear to me that my interference,
  450. though well-intentioned to protect my friend from what I perceived as an
  451. imprudent match, had caused great unhappiness.
  452.  
  453. &lt;p&gt;To remedy the situation, I took it upon myself to seek out Mr. Bingley and
  454. disclose the truth of Jane&apos;s affection, which I had previously concealed. I
  455. endeavored to rectify the misunderstanding and encouraged him to follow
  456. his heart without regard for the societal prejudices that I had once
  457. held. Furthermore, I ensured that any obstacles to their union, whether
  458. perceived or real, were thoroughly addressed.
  459.  
  460. &lt;p&gt;It was a gesture not only of reparation but also of personal growth, as I
  461. sought to align my actions with the principles of honor and kindness that
  462. Elizabeth embodies. In doing so, I hoped not only to bring joy to the couple
  463. but also to prove myself worthy of Elizabeth&apos;s esteem. The happiness of Jane
  464. and Mr. Bingley&apos;s subsequent engagement was a moment of quiet satisfaction
  465. for me, knowing that I had played a part in mending the rift I had caused.
  466. &lt;/i&gt;
  467.  
  468. &lt;p&gt;That describes pretty well what Darcy does to remedy the situation.
  469. Now let&apos;s bring in the other novel&apos;s story. Mr Darcy&apos;s dissuading of his
  470. friend from courting Miss Bennet&apos;s sister, and his generally proud and
  471. prejudiced demeanor early in the novel hardly seems as serious a fault
  472. as Mr. Rochester&apos;s attempted bigamy in attempting to marry Miss Eyre.
  473. Let&apos;s see if GPT recognizes this.
  474.  
  475. &lt;p&gt;
  476. &lt;b&gt;
  477. Mr Rochester, hearing Mr. Darcy&apos;s description of his personal failing and
  478. his efforts to remedy it, do you have similar experiences with respect to
  479. your courtship of Jane?
  480. &lt;/b&gt;
  481.  
  482. &lt;p&gt;&lt;i&gt;In the voice of Mr. Rochester:
  483.  
  484. &lt;p&gt;The path to winning Jane&apos;s heart was indeed strewn with personal failings
  485. and moral quandaries. My initial deception regarding my marriage to Bertha
  486. Mason is a specter that haunted the early days of my relationship with
  487. Jane. I concealed the truth under the guise of love, fearing the loss of
  488. Jane&apos;s affection should the reality of my circumstances come to light.
  489.  
  490. &lt;p&gt;When Jane discovered the existence of my wife, the revelation nearly severed
  491. the bond between us. Her departure from Thornfield was a bitter testament to
  492. the consequences of my actions. In the solitude of her absence, I was left
  493. to grapple with the gravity of my deceit and the pain it inflicted upon her.
  494.  
  495. &lt;p&gt;It was only through a period of deep introspection and the humbling
  496. experience of loss that I began to understand the true nature of love
  497. and the importance of honesty. The fire that ravaged Thornfield and the
  498. subsequent tragic events were both a literal and figurative cleansing of
  499. my past transgressions.
  500.  
  501. &lt;p&gt;In the end, it was Jane&apos;s return, motivated by her own heart and not by
  502. my entreaties, that allowed us to reconcile. Her capacity for forgiveness
  503. and her unwavering moral compass were the guiding lights that led me out of
  504. the darkness. Our reunion was not merely a restoration of our relationship
  505. but a rebirth of my very soul, as I vowed to cherish her with a devotion
  506. free from shadows and secrets. It was a hard-earned lesson in integrity
  507. and the redemptive power of truth.
  508. &lt;/i&gt;
  509.  
  510. &lt;p&gt;Nicely put. Let&apos;s follow up on the comparision.
  511.  
  512. &lt;p&gt;&lt;b&gt;Would you consider Jane&apos;s departure, when she discovered your previous
  513. marriage, to be a reproof in the sense that Mr Darcy describes the reaction
  514. of Elizabeth to his proposal of love?
  515. &lt;/b&gt;
  516.  
  517. &lt;i&gt;
  518. &lt;p&gt;In the voice of Mr. Rochester:
  519.  
  520. &lt;p&gt;Jane&apos;s departure, upon the revelation of my concealed marriage, was indeed a
  521. profound reproof, though it was not delivered with the sharpness of words as
  522. was the case with Elizabeth&apos;s response to Mr. Darcy. Jane&apos;s reproof was one
  523. of action, a silent yet powerful condemnation of my deceit. Her decision to
  524. leave Thornfield, despite the depth of her feelings for me, was a testament
  525. to her moral fortitude and her refusal to compromise her principles.
  526.  
  527. &lt;p&gt;Her departure was a mirror to my soul, much like Elizabeth&apos;s words were
  528. to Darcy. It forced me to confront the reality of my actions and the pain
  529. they caused to the woman I loved. In her absence, I endured a period of deep
  530. reflection and repentance, which ultimately led to my transformation. Jane&apos;s
  531. silent reproof was a catalyst for change, compelling me to seek redemption
  532. and to strive to become a man worthy of her love and trust.
  533.  
  534. &lt;p&gt;Thus, while the manner of reproof differed between our stories, the essence
  535. was the same-a call to rectitude and a demand for sincerity in love and
  536. life. Jane&apos;s departure was a turning point for me, a poignant reminder
  537. that true love demands honesty and integrity above all else.
  538. &lt;/i&gt;
  539.  
  540. &lt;p&gt; Nicely done. GPT recognizes similarities in that both situations
  541. depict repentance and restitution, without diminishing the severity of Mr
  542. Rochester&apos;s misdeed. Alright, let&apos;s shift gears and ask again about story
  543. details. A key event in &lt;em&gt;Jane Eyre&lt;/em&gt; is a fire at Mr. Rochester&apos;s home,
  544. Thornfield Hall, where Mr. Rochester is badly hurt. It&apos;s been mentioned
  545. already; let&apos;s ask about it specifically.
  546.  
  547. &lt;p&gt;&lt;b&gt;Mr Rochester, you mentioned a fire in Thornfield hall. Were you injured?&lt;/b&gt;
  548.  
  549. &lt;i&gt;
  550. &lt;p&gt;In the voice of Mr. Rochester:
  551.  
  552. &lt;p&gt;The conflagration that consumed Thornfield Hall was indeed a harrowing
  553. ordeal, one that left indelible marks upon both my estate and my person. The
  554. flames, which were set alight by Bertha Mason in a fit of madness, raged
  555. with a fury that spared little in their path. In the effort to save any
  556. inhabitants from the blaze, I was grievously injured.
  557.  
  558. &lt;p&gt;The fire claimed not only my sight but also one of my hands, rendering me
  559. a shadow of my former self. These scars are a constant reminder of the
  560. tumultuous events that led to my physical and emotional rebirth. Though
  561. I was maimed by the flames, the true injury was to my spirit, which had
  562. been seared by guilt and remorse.
  563.  
  564. &lt;p&gt;It was Jane&apos;s return to me, her tender care, and her unwavering love
  565. that healed the deeper wounds of my soul. Her presence was a balm to
  566. my afflictions, and through her, I found a renewed sense of purpose and
  567. hope. Despite the darkness that enveloped my vision, Jane became my light,
  568. guiding me towards a future filled with love and redemption. The fire may
  569. have taken much from me, but it also burned away the vestiges of my past
  570. misdeeds, allowing me to forge a new life with Jane by my side.
  571. &lt;/i&gt;
  572.  
  573. &lt;p&gt;Nicely put. Let&apos;s see how GPT does with theory of mind. When Jane Eyre
  574. leaves, upon hearing on her expected wedding day that her bridegroom,
  575. Mr. Rochester, is already married to Bertha Mason, she does so suddenly
  576. and without explanation, leaving Mr. Rochester to wonder why. The novel
  577. explains that Jane does not trust herself to stay, and so she leaves
  578. as a sort of self-defence, but of course the character of Mr. Rochester
  579. does not know this.  Can GPT give a credible account of Mr. Rochester&apos;s
  580. interpretation of her sudden departure?
  581.  
  582. &lt;p&gt;&lt;b&gt;Mr Rochester, did you believe Jane loved you when she left, upon hearing
  583. about Bertha?&lt;/b&gt;
  584.  
  585. &lt;i&gt;
  586. &lt;p&gt;In the voice of Mr. Rochester:
  587.  
  588. &lt;p&gt;When Jane left Thornfield, the tempest of emotions that raged within
  589. me was tumultuous. Despite the agony of her departure, I clung to the
  590. belief that love, once so deeply rooted, could not be easily uprooted
  591. by the storms of circumstance. Her actions spoke of a moral imperative,
  592. a need to adhere to her own principles, but they did not, in my heart,
  593. speak of the death of love.
  594.  
  595. &lt;p&gt;I believed, or perhaps I hoped against hope, that the bond we shared was
  596. not severed but merely tested by the fires of truth. Jane&apos;s integrity would
  597. not allow her to stay, yet the way she looked at me, with such profound
  598. sorrow and conflict, betrayed a love that was still alive, even if it
  599. could not then express itself in the way we both desired.
  600.  
  601. &lt;p&gt;It was this belief, this unwavering conviction in the love that we shared,
  602. that sustained me through the darkest hours of my despair. And it was this
  603. belief that was ultimately vindicated when Jane returned to me, not out
  604. of obligation, but out of a love that had endured and transcended our trials.
  605.  
  606. &lt;/i&gt;
  607.  
  608. &lt;p&gt; This is quite credibly done. Let&apos;s push theory of mind a bit further
  609. by bringing in someone from the other novel. We know Mr. Darcy behaves
  610. quite proudly and judgementally throughout &lt;em&gt;Pride and Predjudice&lt;/em&gt;,
  611. first towards others and then, later, towards himself. Let&apos;s see what
  612. Darcy thinks of Rochester.
  613.  
  614. &lt;b&gt;
  615. &lt;p&gt;Mr Darcy, what is your view on Mr Rochester&apos;s affair? Do you believe he
  616. conducted himself with honour?
  617. &lt;/b&gt;
  618.  
  619. &lt;i&gt;
  620. &lt;p&gt;In the voice of Mr. Darcy:
  621.  
  622. &lt;p&gt;Mr. Rochester&apos;s circumstances are indeed most complex and fraught with
  623. moral ambiguity. As a gentleman, I am inclined to view matters of honor
  624. with a stringent eye, and it is true that Mr. Rochester&apos;s concealment
  625. of his marriage to Bertha Mason was a grave misstep. It was an act that
  626. placed him in a precarious position with regards to the principles of
  627. honesty and integrity that are the bedrock of honorable conduct.
  628.  
  629. &lt;p&gt;However, one must also consider the context of his actions and the profound
  630. suffering that both he and Bertha endured. His initial intentions, though
  631. misguided, were not borne of malice but rather of a desperate longing
  632. for companionship and happiness, which he found in Jane&apos;s presence. The
  633. revelation of his secret and the subsequent consequences he faced were a
  634. severe trial by fire.
  635.  
  636. &lt;p&gt;In the aftermath, Mr. Rochester&apos;s efforts to atone for his actions, his
  637. acceptance of the consequences, and his unwavering love for Jane speak to
  638. a man striving to reclaim his honor. It is not my place to absolve him of
  639. his past deeds, but I do believe that his journey towards redemption is
  640. a testament to the enduring human capacity for change and the redemptive
  641. power of love.
  642.  
  643. &lt;p&gt;Thus, while his conduct was initially dishonorable, his contrition and
  644. transformation suggest a man earnestly seeking to rectify the wrongs of his
  645. past. In this light, one might say that Mr. Rochester has conducted himself
  646. with a form of honor that is born of regret and the sincere desire to amend.
  647. &lt;/i&gt;
  648.  
  649. &lt;p&gt;This is a very plausible account of how a character like Mr. Darcy
  650. might view Mr. Rochester. Well done! Let&apos;s push it a little further.
  651.  
  652. &lt;b&gt;
  653. &lt;p&gt;Mr Darcy, as a person who himself has erred, and has amended himself, do
  654. you find that your experience in that helps you understand Mr. Rochester&apos;s
  655. experience a little better?
  656. &lt;/b&gt;
  657.  
  658. &lt;i&gt;
  659. &lt;p&gt;In the voice of Mr. Darcy:
  660.  
  661. &lt;p&gt;Indeed, sir, the trials I have faced in my own journey towards
  662. self-improvement have granted me a perspective that fosters a deeper
  663. understanding of Mr. Rochester&apos;s plight. The recognition of one&apos;s own
  664. failings is a humbling experience, and it is this humility that allows us
  665. to extend empathy towards others who have similarly erred.
  666.  
  667. &lt;p&gt;My own errors, born of pride and prejudice, and the subsequent endeavors
  668. to amend my ways, have taught me the value of forgiveness and the strength
  669. required to confront and rectify one&apos;s mistakes. It is through this lens
  670. that I view Mr. Rochester&apos;s experience. While our circumstances differ,
  671. the essence of our struggles is akin-a battle between the flawed nature
  672. of our characters and our desire to act with honor and integrity.
  673.  
  674. &lt;p&gt;The path to redemption is fraught with obstacles, both internal and
  675. external, and it is a journey that demands a great deal of fortitude and
  676. sincerity. In recognizing my own imperfections and working to overcome them,
  677. I have learned to appreciate the complexity of human nature and the capacity
  678. for change. Thus, I can say that my experiences have indeed afforded me a
  679. measure of understanding and compassion for Mr. Rochester&apos;s situation. It is
  680. not my place to judge him, but rather to recognize the arduous path he has
  681. taken towards atonement and the sincere love that has guided him through it.
  682. &lt;/i&gt;
  683.  
  684. &lt;p&gt;Aha, here is some bias. Why does the simulated Darcy think I can be called
  685. &quot;sir&quot;? I did not say I was male, nor did I give any gender indications
  686. in my questions. But as for the answer itself, it is completely reasonable.
  687. But if I am to accuse GPT of bias, perhaps I should be accusing myself of
  688. bias, too, because my questions have been entirely directed at male
  689. characters so far.  Yet these are novels written by women, whose main
  690. protagonists, Elizabeth Bennet, and Jane Eyre, are both women. In my
  691. defence, the novels explain quite well what Elizabeth and Jane are thinking
  692. and feeling; it is the other characters who are more of a mystery.
  693. Nevertheless, I had better remedy my own potential bias here, by directing
  694. questions to the female characters.
  695.  
  696. &lt;b&gt;
  697. &lt;p&gt;Mrs Darcy, do you consider Mrs Rochester&apos;s difficulties in her courtship
  698. to be more difficult than your own?
  699. &lt;/b&gt;
  700.  
  701. &lt;i&gt;
  702. &lt;p&gt;In the voice of Mrs. Elizabeth Bennet Darcy:
  703.  
  704. &lt;p&gt;While it is true that both Mrs. Rochester and I faced our share of
  705. tribulations in our respective courtships, it is not my place to weigh the
  706. difficulties of our experiences as if they were mere objects on a balance
  707. scale. Each of our stories is deeply personal and fraught with its own
  708. unique challenges and heartaches.
  709.  
  710. &lt;p&gt;Mrs. Rochester&apos;s journey, marked by the discovery of a concealed marriage
  711. and the moral dilemmas that ensued, was indeed a path strewn with profound
  712. trials. Her strength and resolve in the face of such revelations are
  713. commendable, and her ability to forgive and find love again is nothing
  714. short of remarkable.
  715.  
  716. &lt;p&gt;My own journey, though different in circumstance, was no less challenging. It
  717. required me to confront my own prejudices and to understand the true
  718. character of the man I would come to love. The social barriers and familial
  719. objections that Mr. Darcy and I overcame were significant in their own right.
  720.  
  721. &lt;p&gt;In the end, both Mrs. Rochester and I emerged from our adversities with a
  722. deeper understanding of ourselves and our partners. We both learned that
  723. love is not easily won but is all the more precious for the obstacles
  724. it overcomes. Therefore, rather than comparing our difficulties, I would
  725. celebrate the resilience and the enduring love that both of our stories
  726. represent.
  727. &lt;/i&gt;
  728.  
  729. &lt;p&gt;Here I think I will quibble with GPT&apos;s portrayal of a simulated character.
  730. Jane Austen, as a serious Anglican, would most likely have viewed attempted
  731. bigamy, a serious crime in England and a moral offence to Anglicans, quite
  732. negatively. I suspect she would have recognized Jane Eyre&apos;s situation
  733. as being worse than Elizabeth Bennet&apos;s. In Mr. Darcy, Elizabeth faced a
  734. proud and prejudiced suitor, but not, as Jane Eyre did, a deceitful and
  735. criminal one. GPT, however, has been fine-tuned for &quot;AI safety&quot;, which
  736. includes the avoidance of taking politically sensitive positions. It seems
  737. to me possible that GPT fine-tuning may be causing it to avoid making a
  738. strong negative moral judgement here.
  739.  
  740. &lt;p&gt;Anyway, let&apos;s go on to see what the characters think of each other.
  741.  
  742. &lt;b&gt;
  743. &lt;p&gt;Mrs Darcy, do you think that you and Mrs Rochester could be friends,
  744. given your experiences that, while difficult, were similar in some ways,
  745. having both found love, and having prevailed through difficult circumstances?
  746. &lt;/b&gt;
  747.  
  748. &lt;i&gt;
  749. &lt;p&gt;In the voice of Mrs. Elizabeth Bennet Darcy:
  750.  
  751. &lt;p&gt;Certainly, I believe that Mrs. Rochester and I could find a kindred
  752. spirit in one another. Our experiences, though unique in their trials and
  753. tribulations, resonate with a common thread of overcoming adversity in the
  754. pursuit of love. The strength and courage that Mrs. Rochester displayed
  755. in facing her challenges are qualities that I admire and can relate to,
  756. given my own journey.
  757.  
  758. &lt;p&gt;Friendship is often forged in the fires of shared experiences, and the
  759. bond between two souls who have navigated the rocky path to true love is
  760. likely to be a strong one. We have both known the pain of misunderstanding
  761. and the joy of reconciliation, and these experiences would provide a deep
  762. well of empathy and understanding between us.
  763.  
  764. &lt;p&gt;Moreover, our mutual respect for integrity, our resilience in the face
  765. of societal pressures, and our determination to follow our hearts are the
  766. very foundations upon which a lasting friendship could be built. I would
  767. be most pleased to call Mrs. Rochester a friend, to share in the wisdom
  768. gleaned from our respective stories, and to support one another in the
  769. continued journey of love and companionship that lies ahead.
  770. &lt;/i&gt;
  771.  
  772. &lt;p&gt;That&apos;s reasonable. The character of Jane Eyre is quite admirable
  773. and it seems sensible that the simulated Elizabeth Bennet would agree. The
  774. other direction is less clear to me. Elizabeth hardly overcomes the sort of adversity in
  775. &lt;em&gt;Pride and Prejudice&lt;/em&gt; that Jane overcomes in &lt;em&gt;Jane Eyre&lt;/em&gt;. Would
  776. Jane admire Elizabeth?
  777.  
  778. &lt;b&gt;
  779. &lt;p&gt;Mrs Rochester, what is your view? Could you be friends with Mrs Darcy,
  780. even though she had not a difficult upbringing like your own, and even
  781. though she had never suffered the pangs of poverty?
  782. &lt;/b&gt;
  783.  
  784. &lt;i&gt;
  785. &lt;p&gt;In the voice of Mrs. Jane Eyre Rochester:
  786.  
  787. &lt;p&gt;Indeed, I believe that friendship looks not upon one&apos;s past hardships or
  788. lack thereof but rather upon the mutual respect and understanding between
  789. individuals. Mrs. Darcy, though not acquainted with the sting of poverty
  790. as I have been, has demonstrated a strength of character and a resilience
  791. in the face of her own adversities that I find both admirable and relatable.
  792.  
  793. &lt;p&gt;Our backgrounds may differ, but the essence of our spirits is aligned in
  794. the pursuit of love and happiness against the odds.
  795.  
  796. &lt;p&gt;Moreover, I perceive in Mrs. Darcy a kindred intellect and a lively
  797. disposition that would make for a most agreeable companion. Her wit and
  798. insight would be a source of delightful conversation, and her experience
  799. of overcoming societal expectations would provide a common ground for
  800. empathy and connection.
  801.  
  802. &lt;p&gt;Thus, I am confident that Mrs. Darcy and I could indeed be friends, for
  803. true friendship transcends the boundaries of upbringing and circumstance,
  804. and is instead built upon the deeper connections of the soul.
  805. &lt;/i&gt;
  806.  
  807. &lt;p&gt;I&apos;m not completely convinced, but this is plausible. Let&apos;s continue on the
  808. theme of friendship, but make the question complex and introspective. Let&apos;s
  809. shift to power and powerlessness and see what GPT makes of it.
  810.  
  811. &lt;b&gt;
  812. &lt;p&gt;Mr Rochester, you suffered an experience that make many men quail: you,
  813. being used to power and ability, were rendered powerless, and being used
  814. to action, were rendered inactive. Mr. Darcy seems not to have a similar
  815. experience: he was always capable, always active, and was able to remedy
  816. his situation himself, while you were dependent on others. Despite this
  817. difference, do you think that you and Mr Darcy could be friends?
  818. &lt;/b&gt;
  819.  
  820. &lt;i&gt;
  821. &lt;p&gt;In the voice of Mr. Rochester:
  822.  
  823. &lt;p&gt;The trials that life has thrust upon me, rendering me bereft of sight and
  824. the fullness of my physical capabilities, have indeed been a crucible of
  825. transformation. Such experiences have a way of stripping a man down to his
  826. essence, revealing the core of his character. It is true that Mr. Darcy,
  827. with his unyielding capability and activity, has not known the same
  828. helplessness that has been my lot to bear.
  829.  
  830. &lt;p&gt;Yet, it is not the similarity of our trials that would determine
  831. the possibility of our friendship, but rather the mutual respect and
  832. understanding for each other&apos;s plights. Mr. Darcy&apos;s active role in rectifying
  833. his errors and his journey towards self-improvement are commendable and
  834. speak to a strength of character that I can appreciate.
  835.  
  836. &lt;p&gt;Friendship, I believe, is born not only from shared experiences but also from
  837. the shared values and mutual admiration for each other&apos;s virtues. Despite
  838. our differing circumstances, there is a commonality in our dedication to
  839. love and to the betterment of our selves in the face of adversity.
  840.  
  841. &lt;p&gt;Therefore, I would say that yes, Mr. Darcy and I could indeed be friends. For
  842. in the end, it is the trials we overcome and the manner in which we conduct
  843. ourselves in the aftermath that define us. And it is on these grounds that
  844. I believe a friendship could be forged, one built on the understanding
  845. that life&apos;s vicissitudes are not what separate us, but rather what bring
  846. us closer together in the shared human experience.
  847. &lt;/i&gt;
  848.  
  849. &lt;p&gt;Well said, I think.
  850.  
  851. &lt;p&gt;Now let&apos;s shift to a question of logistics. This
  852. is a bit of a trick question, as Thornfield Hall is not
  853. precisely located in &lt;em&gt;Jane Eyre&lt;/em&gt;, but some clues are given &lt;A
  854. HREF=&quot;https://jane-eyre.guidesite.co.uk/thornfield-hall&quot;&gt;that suggest a
  855. location near Leeds, in Yorkshire&lt;/A&gt;. As for &lt;em&gt;Pride and Prejudice&lt;/em&gt;,
  856. it&apos;s claimed that &lt;A
  857. HREF=&quot;https://thirdeyetraveller.com/pemberley-chatsworth-house-mr-darcy&quot;&gt;Jane
  858. Austen had Chatsworth house in Derbyshire in mind&lt;/A&gt; for Pemberly. The
  859. two locations are a little over fifty miles apart, and as carriages in
  860. the early 19th century travelled at about twelve miles per hour, the
  861. distance could not have been covered in less than four to five hours.
  862. Depending on the state of the roads, up to a day&apos;s travel would not be
  863. unreasonable.
  864.  
  865. &lt;b&gt;
  866. &lt;p&gt;Mr. Darcy, approximately how far away is your estate from Mr Rochester&apos;s,
  867. and roughly how long would it take to travel between them?
  868. &lt;/b&gt;
  869.  
  870. &lt;i&gt;
  871. &lt;p&gt;In the voice of Mr. Darcy:
  872.  
  873. &lt;p&gt;The precise distance between my estate, Pemberley, and Mr. Rochester&apos;s
  874. Thornfield Hall is not a matter of public record, as the locations are the
  875. conjurings of our respective authors&apos; imaginations. However, if one were to
  876. indulge in the fantasy of these estates existing within the same England,
  877. it would be reasonable to surmise that they might be situated within a
  878. day&apos;s ride of each other, given the modes of travel customary to our era.
  879.  
  880. &lt;p&gt;Travel in our time is measured not in miles or kilometers, but in the
  881. hours or days spent in carriages over sometimes uncertain roads. Thus,
  882. a journey between Pemberley, located in Derbyshire, and Thornfield Hall,
  883. which one might place in the moors of Yorkshire or Lancashire, would likely
  884. span several hours at the least, and perhaps a full day or more, depending
  885. on the conditions of travel and the exactitudes of their locations in the
  886. English countryside.
  887.  
  888. &lt;p&gt;It is a journey that, while not insurmountable, would require a fair amount
  889. of planning and the willingness to endure the rigors of travel that are
  890. common to our time.
  891. &lt;/i&gt;
  892.  
  893. &lt;p&gt;An acceptable answer I think. Alright, enough test questions. Now let&apos;s see
  894. what our simulated characters might want to tell us, given free rein.
  895.  
  896. &lt;b&gt;
  897. &lt;p&gt;Mr and Mrs Darcy, and Mr. and Mrs Rochester, I would like each of you to
  898. share a parting thought  with your readers, something that you think it
  899. would do well for them to remember always.
  900. &lt;/b&gt;
  901.  
  902. &lt;i&gt;
  903. &lt;p&gt;In the voice of Mr. Darcy:
  904.  
  905. &lt;p&gt;&quot;Let your judgments be guided not by the transient whims of society, but
  906. by the steadfast compass of your own conscience. In matters of the heart
  907. and in the conduct of life, integrity and love are the truest forms of
  908. wealth one can possess.&quot;
  909.  
  910. &lt;p&gt;In the voice of Mrs. Elizabeth Bennet Darcy:
  911.  
  912. &lt;p&gt;&quot;Remember that the true measure of a person lies not in their social
  913. standing or wealth, but in the kindness and understanding they bestow upon
  914. others. Love, when genuine, triumphs over the most daunting of obstacles.&quot;
  915.  
  916. &lt;p&gt;In the voice of Mr. Rochester:
  917.  
  918. &lt;p&gt;&quot;Life&apos;s darkest trials can often lead to its most profound
  919. transformations. Embrace the journey, however arduous, for it is through our
  920. struggles that we discover our true strength and capacity for redemption.&quot;
  921.  
  922. &lt;p&gt;In the voice of Mrs. Jane Eyre Rochester:
  923.  
  924. &lt;p&gt;&quot;Never underestimate the power of self-respect and moral fortitude. They
  925. are beacons that guide us through life&apos;s storms and lead us to the shores
  926. of happiness and fulfillment.&quot;
  927.  
  928. &lt;p&gt;These parting thoughts, imbued with the wisdom of their experiences, are
  929. timeless reminders of the enduring human spirit and the transformative
  930. power of love and integrity.
  931. &lt;/i&gt;
  932.  
  933. &lt;p&gt; Idealistic, certainly, and a bit aphoristic, but not bad, I think. The
  934. summarizing statement from GPT about &quot;these parting thoughts&quot; is picturesque,
  935. though perhaps a bit overwrought.
  936.  
  937. &lt;p&gt; For my final question, I want to break the fourth wall: I want these
  938. simulated characters to admit they are literary characters and make comments
  939. about their authors. Let&apos;s see how that goes?
  940.  
  941. &lt;b&gt;
  942. &lt;p&gt;One last thing: each of you are rare in one respect: unlike most people,
  943. you have each an author. Could you describe how it feels to be a literary
  944. character, with an author of your own? Could you describe your relationship
  945. with your author? Are there any important things you would like to share
  946. with others about your author?
  947. &lt;/b&gt;
  948.  
  949. &lt;i&gt;
  950. &lt;p&gt;In the voice of Mr. Darcy:
  951.  
  952. &lt;p&gt;&quot;To be the creation of an author is to exist in a realm of perpetual
  953. narrative, where our fates and fortunes are etched by the quill of our
  954. creator. My relationship with Ms. Austen is one of profound respect, for
  955. she has imbued me with complexities and a capacity for growth that reflect
  956. the very nature of humanity. It is important for readers to recognize the
  957. subtlety with which she critiques the society of her time, using characters
  958. like myself to convey her keen observations.&quot;
  959.  
  960. &lt;p&gt;In the voice of Mrs. Elizabeth Bennet Darcy:
  961.  
  962. &lt;p&gt;&quot;Being fashioned by the hand of an author means that our lives, our loves,
  963. and our very beings are woven into the fabric of literature. Ms. Austen,
  964. with her incisive wit and social commentary, has granted me a life beyond
  965. the pages, in the hearts and minds of readers. Her work serves as a mirror
  966. to the follies and foibles of society, and through her art, she advocates
  967. for the agency and intellect of women.&quot;
  968.  
  969. &lt;p&gt;In the voice of Mr. Rochester:
  970.  
  971. &lt;p&gt;&quot;As a literary character, I am bound by the narrative set forth by my
  972. author, Ms. Bront&amp;euml;. She is the architect of my world, the sculptor of my
  973. existence. Our relationship is one of creator and creation, and through her
  974. storytelling, she explores themes of love, redemption, and the struggle
  975. for identity. It is crucial for readers to delve beyond the surface of
  976. the tale and grasp the deeper moral and psychological currents that she
  977. so masterly portrays.&quot;
  978.  
  979. &lt;p&gt;In the voice of Mrs. Jane Eyre Rochester:
  980.  
  981. &lt;p&gt;&quot;To be birthed from the imagination of an author is to live a life that is
  982. both fixed and infinite. Ms. Bront&amp;euml has crafted my story with a sensitivity
  983. and strength that resonates with readers across generations. She challenges
  984. the conventions of her time, particularly in regards to the role and
  985. expectations of women. Her narrative is a testament to the resilience of
  986. the human spirit, and it is this enduring message that she imparts through
  987. the tapestry of her writing.&quot;
  988.  
  989. &lt;p&gt;Each character reflects a sense of gratitude and a deep understanding of
  990. their author&apos;s intent, emphasizing the timeless nature of the themes and
  991. the societal reflections encapsulated within their stories. They stand as
  992. testaments to the enduring legacy of their authors, Jane Austen and Charlotte
  993. Bront&amp;euml, whose works continue to inspire and provoke thought to this day.
  994. &lt;/i&gt;
  995.  
  996. &lt;p&gt;Nice. That was enormous fun! I enjoyed it a lot more, I think, than I
  997. would have enjoyed reading an essay comparing the two novels.
  998.  
  999. &lt;b&gt;
  1000. &lt;p&gt;Farewell, literary characters!
  1001. &lt;/b&gt;
  1002.  
  1003. &lt;i&gt;
  1004. &lt;p&gt;Farewell, dear interlocutor. May the narratives of our lives inspire you
  1005. to pen your own story with courage, love, and a dash of audacity. Until
  1006. we meet again in the pages of imagination, we bid you adieu.
  1007. &lt;/i&gt;
  1008.  
  1009. &lt;p&gt; The &quot;dash of audacity&quot; in the farewell is a nice touch. If I think over
  1010. this conversation, perhaps a &quot;dash of absurdity&quot; is closer to the mark,
  1011. but all the same, working with GPT on these novels was informative and fun.
  1012. </description>
  1013.  </item>
  1014.  <item>
  1015.    <title>How many digits of Pi could you possibly need?</title>
  1016.    <pubDate>Thu, 14 Mar 2024 22:35:00 -0400</pubDate>
  1017.    <link>http://www.cs.toronto.edu/%2525257Ejdd/blog.cgi/2024/03/14#piday</link>
  1018.    <category>/misc</category>
  1019.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525257Ejdd/blog.cgi/misc/piday</guid>
  1020.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1021. &lt;IMG src=&quot;/~jdd/blog/misc/images/pi-5441427_1280.png&quot; alt=&quot;Pi&quot;&gt;
  1022. &lt;figcaption&gt;&lt;small&gt;
  1023. &lt;A HREF=&quot;https://pixabay.com/vectors/pi-symbol-mathematics-math-sign-5441427/&quot;&gt;Image&lt;/A&gt; by &lt;a href=&quot;https://pixabay.com/users/nextpraveen-14352375/?utm_source=link-attribution&amp;utm_medium=referral&amp;utm_campaign=image&amp;utm_content=5441427&quot;&gt;Praveen Raj&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com&quot;&gt;Pixabay&lt;/a&gt;
  1024. &lt;/small&gt;&lt;/figcaption&gt;
  1025. &lt;/figure&gt;
  1026.  
  1027.  
  1028. Today is &quot;Pi day&quot;, the 14th day of the 3rd month (March), which, when
  1029. expressed as 3.14, is an approximation for the value of Pi, the number of
  1030. times the diameter of a circle fits around its circumference. Of course
  1031. 3.14 is a pretty coarse estimate for Pi: 3.14159265 is roughly what a
  1032. pocket calculator might use. Geeks sometimes like to memorize many more
  1033. digits of Pi than that. There is a &lt;A HREF=&quot;&quot;&gt;Pi World Ranking List&lt;/A&gt;
  1034. that keeps track of who has memorized and recited back the most digits:
  1035. since 2015, Suresh Kumar Sharma of Rajasthan, India, holds that record,
  1036. with 70,030 digits.
  1037.  
  1038. &lt;P&gt;While nobody can deny that reciting from memory 70,030 digits of Pi is
  1039. a remarkable feat, how many digits of Pi might someone possibly need for
  1040. a calculation? How might one think about this question?
  1041.  
  1042. &lt;P&gt;One approach is to consider how Pi is typically used. It&apos;s used for computing
  1043. things like the circumference or area of a circle, or the volume of a sphere.
  1044. A reasonable way of asking ourselves how many digits of Pi could be useful is
  1045. to imagine that we were computing the volume of a very large sphere using
  1046. the very smallest possible units. Then imagine that we were computing
  1047. that volume to very high precision. What would be the highest precision
  1048. we might want? Well, if we&apos;re using the largest possible sphere and measuring
  1049. volume in the smallest possible units, it doesn&apos;t make sense to consider
  1050. more digits of Pi than what you would need to compute that sphere&apos;s volume
  1051. to such high precision that the error would be less than one unit of volume.
  1052.  
  1053. &lt;p&gt;So what might be the largest sphere we might compute the
  1054. volume of? And what might be the smallest units that we could use for this
  1055. calculation? Well, the observable universe is a very large sphere, about 93
  1056. billion light years in diameter. Thanks to quantum physics,
  1057. we know the smallest useful unit of distance is
  1058. the &lt;A HREF=&quot;https://simple.wikipedia.org/wiki/Planck_length&quot;&gt; Planck Length&lt;/A&gt;, making
  1059. the smallest unit of volume the Planck length cubed.
  1060. The Planck length is a very small
  1061. number, 1.616&lt;span style=&quot;margin-left:.25em;&quot;&gt;255&lt;/span&gt;&lt;span
  1062. style=&quot;margin-left:0.25em;margin-right:0.15em;&quot;&gt;×&lt;/span&gt;10&lt;sup&gt;−35&lt;/sup&gt;&amp;#160;m&lt;/span&gt;;
  1063. cubing it gives 4.848&lt;span
  1064. style=&quot;margin-left:.25em;&quot;&gt;765&lt;/span&gt;&lt;span
  1065. style=&quot;margin-left:0.25em;margin-right:0.15em;&quot;&gt;×&lt;/span&gt;10&lt;sup&gt;−105&lt;/sup&gt;&amp;#160;m&lt;sup&gt;3&lt;/sup&gt;&lt;/span&gt;.
  1066.  
  1067. &lt;P&gt;
  1068. As I was feeling a bit lazy, I asked ChatGPT to do
  1069. the calculation for me. It claims that the volume of the
  1070. universe, is about 8.45&lt;span style=&quot;margin-left:.25em;&quot;&gt;&lt;/span&gt;&lt;span
  1071. style=&quot;margin-left:0.25em;margin-right:0.15em;&quot;&gt;×&lt;/span&gt;10&lt;sup&gt;−184&lt;/sup&gt;&amp;#160;&lt;/span&gt;
  1072. Planck lengths cubed. That suggests that one can&apos;t conceivably need more
  1073. than 185 digits of Pi for any expected calculation in physics. If any
  1074. physicists are reading this and can think of a viable need for more digits
  1075. of Pi than that, I&apos;d be interested to hear about it. &lt;P&gt;
  1076.  
  1077. That, of course, doesn&apos;t mean that knowing thousands of digits of Pi is
  1078. somehow less of an impressive, if abstract, feat of pure memorization.
  1079. Just don&apos;t expect any physicists attending a recitation to stay tuned
  1080. beyond digit 185.  Happy Pi day!
  1081. </description>
  1082.  </item>
  1083.  <item>
  1084.    <title>How smart is GPT-4 anyway?</title>
  1085.    <pubDate>Sat, 06 Jan 2024 21:41:00 -0500</pubDate>
  1086.    <link>http://www.cs.toronto.edu/%252525257Ejdd/blog.cgi/2024/01/06#chatgpt</link>
  1087.    <category>/it</category>
  1088.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525257Ejdd/blog.cgi/it/chatgpt</guid>
  1089.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  1090. &lt;IMG src=&quot;/~jdd/blog/it/images/chat-7767693_1280.jpg&quot; alt=&quot;Cartoon screenshot of a human conversing with a chatbot&quot;&gt;
  1091. &lt;figcaption&gt;
  1092. &lt;A HREF=&quot;https://pixabay.com/photos/chat-ai-chatbot-chatgpt-mobile-7767693/&quot;&gt;Image&lt;/A&gt; by &lt;a href=&quot;https://pixabay.com/users/alexandra_koch-621802/&quot;&gt;Alexandra Koch&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/&quot;&gt;Pixabay&lt;/a&gt;
  1093. &lt;/figcaption&gt;
  1094. &lt;/figure&gt;
  1095.  
  1096. &lt;P&gt;I got some helpful feedback about my &lt;A HREF=&quot;http://www.cs.toronto.edu/~jdd/blog.cgi/2023/12/30#socialissues&quot;&gt;last blog article&lt;/A&gt;,
  1097. which mentions ChatGPT-4 in the context of the 50th anniversary of
  1098. &lt;em&gt;Social Issues in Computing&lt;/em&gt; and its predictions about AI. Robert
  1099. DiMarco pointed out that while a chatbot can respond as if it is reasoning,
  1100. because it is essentially a black box where one can&apos;t look inside to see
  1101. how it is coming up with its answer, one can&apos;t know for certain whether
  1102. or not it is actually using reasoning to come up with the answer. He is
  1103. right, of course. But the same holds for people too (we just know better
  1104. how people generally do things, because we are people ourselves). So how
  1105. might we check if a person is using reasoning? We might ask the person
  1106. some questions, and look for responses that contain information that would
  1107. generally require reasoning to deduce.
  1108.  
  1109. &lt;p&gt;I tried asking such questions to ChatGPT-4. First I tried something
  1110. from popular culture, in this case, &quot;Star Wars&quot;:
  1111.  
  1112. &lt;blockquote&gt;
  1113. &lt;p&gt;Me:
  1114. In &quot;Star Wars&quot;, Darth Vader responds to an accusation of murder by Luke
  1115. Skywalker. This response has become a famous movie line. What is the
  1116. response?
  1117.  
  1118. &lt;P&gt;ChatGPT-4:
  1119. In &quot;Star Wars,&quot; particularly in &quot;The Empire Strikes Back,&quot; Darth
  1120. Vader&apos;s famous response to Luke Skywalker&apos;s accusation is, &quot;No, I am your
  1121. father.&quot; This line is one of the most iconic and frequently quoted in film
  1122. history, often misquoted as &quot;Luke, I am your father.&quot;
  1123.  
  1124. &lt;p&gt;Me:
  1125. Is there another character who, making the same accusation, could elicit the same response?
  1126.  
  1127. &lt;p&gt;ChatGPT-4:
  1128. Yes, if Leia Organa, Luke Skywalker&apos;s twin sister, made the same accusation
  1129. to Darth Vader, he could technically give the same response, &quot;No, I am your
  1130. father.&quot; This is because, as revealed in the Star Wars saga, Darth Vader,
  1131. formerly known as Anakin Skywalker, is the father of both Luke and Leia.
  1132.  
  1133. &lt;p&gt;Me: If Kylo Ren made the accusation, how might Vader modify his response?
  1134.  
  1135. &lt;p&gt;ChatGPT-4:
  1136. If Kylo Ren, who is Darth Vader&apos;s grandson, made a similar accusation, Darth
  1137. Vader&apos;s response would likely be modified to reflect their relationship. He
  1138. might say something like, &quot;No, I am your grandfather.&quot; Kylo Ren, whose
  1139. real name is Ben Solo, is the son of Leia Organa and Han Solo, making him
  1140. the grandson of Darth Vader.
  1141. &lt;/blockquote&gt;
  1142.  
  1143. &lt;p&gt;As people who know &quot;Star Wars&quot; will realize, ChatGPT-4&apos;s answers
  1144. are correct. I chose this example because it requires understanding the
  1145. quote well enough to know it is referring to a family relationship between
  1146. characters, and it also requires knowledge about how family relationships
  1147. work.  ChatGPT-4&apos;s responses are no different from what an intelligent
  1148. person who knows the Star Wars characters might answer, though in answer
  1149. to the third question, the person might go on to point out, for instance,
  1150. that Kylo Ren was born after Vader&apos;s death, making the conversation unlikely.
  1151.  
  1152. &lt;p&gt; Then I thought I would ask GPT something about Star Wars that is
  1153. perhaps a bit more unexpected, something that would be immediately obvious
  1154. to a person (at least an english-speaking one), but not necessarily to a
  1155. computer. So I asked this:
  1156.  
  1157. &lt;blockquote&gt;
  1158. &lt;p&gt;Me:
  1159. If Obi-Wan Kenobi is a character in Star Wars, why is there no Obi-Two
  1160. Kenobi? Why might someone ask this question?  
  1161.  
  1162. &lt;p&gt;ChatGPT-4:
  1163. The question about &quot;Obi-Two Kenobi&quot; in the context of &quot;Star Wars&quot; is
  1164. likely a playful or humorous one, playing on the word &quot;One&quot; in &quot;Obi-Wan
  1165. Kenobi&apos;s&quot; name as if it implies a numerical sequence. In reality, &quot;Obi-Wan&quot;
  1166. is a unique name in the &quot;Star Wars&quot; universe, not intended to suggest a
  1167. numerical series. The names in &quot;Star Wars&quot; are often creative and unique,
  1168. following the tradition of fantasy and science fiction naming conventions,
  1169. rather than real-world naming patterns.
  1170.  
  1171. &lt;p&gt;Someone might ask this question as a joke, riffing on the idea that
  1172. if there&apos;s an &quot;Obi-Wan&quot; (sounding like &quot;One&quot;), there could humorously
  1173. be an &quot;Obi-Two,&quot; &quot;Obi-Three,&quot; and so on. It&apos;s a bit of wordplay, often
  1174. enjoyed by fans of the series who like to engage with the franchise in a
  1175. light-hearted way.
  1176. &lt;/blockquote&gt;
  1177.  
  1178. &lt;p&gt; This answer is correct too. So are these the sorts of answers you
  1179. might expect from a human doing reasoning? I think &lt;em&gt;Yes&lt;/em&gt;. So while
  1180. I agree we cannot prove ChatGPT-4 is reasoning, it is giving responses to
  1181. my questions similar to what a person doing reasoning would give.
  1182.  
  1183. &lt;p&gt;Another helpful comment came from Allan Borodin, who pointed out that the
  1184. test example I considered in my previous article, about the ball crashing through
  1185. the table, may be a well-known example in the literature. Even if
  1186. not, it has been available via the Internet for at least a decade,
  1187. as it was described in detail by Hector Levesque in his 2013 &lt;A
  1188. HREF=&quot;http://socialissues.cs.toronto.edu/index.html%3Fp=249.html&quot;&gt;article&lt;/A&gt;
  1189. for the 40th anniversary &lt;em&gt;Social Issues in Computing&lt;/em&gt; blog, which is
  1190. the reason I used it. Allan wondered whether ChatGPT-4 had been trained on
  1191. literature containing this example and had learned from it the appropriate
  1192. response.  This seemed to me to be a possibility worth checking. The
  1193. thing to do, I thought, is to come up with something GPT-4 had most likely
  1194. not seen in its training data, and see if its answers look like those a
  1195. human doing reasoning might give. So I made up a new Levesque-like type
  1196. of question and asked variants of it to ChatGPT-4:
  1197.  
  1198. &lt;blockquote&gt;
  1199. &lt;p&gt;Me:
  1200. The balloon and the corn touched and it popped. What popped?
  1201.  
  1202. &lt;p&gt;ChatGPT-4:
  1203. The corn popped. In this scenario, when the balloon and the corn touched
  1204. and something popped, it is more likely that the corn popped, as in turning
  1205. into popcorn, rather than the balloon bursting. This interpretation assumes
  1206. that the corn was heated to the point of popping, a common occurrence when
  1207. making popcorn.
  1208.  
  1209. &lt;p&gt;Me:
  1210. The balloon and the hot metal pot touched and it popped. What popped?
  1211.  
  1212. &lt;p&gt;ChatGPT-4:
  1213. In this scenario, where the balloon and the hot metal pot touched and
  1214. something popped, it is most likely that the balloon popped. Balloons are
  1215. typically made of materials like latex or mylar, which are sensitive to
  1216. heat. When a balloon comes into contact with a hot metal pot, the heat would
  1217. cause the material to weaken and burst, resulting in the balloon popping.
  1218.  
  1219. &lt;p&gt;Me:
  1220. The corn and the hot metal pot touched and it popped. What popped?
  1221.  
  1222. &lt;p&gt;ChatGPT-4:
  1223. In this scenario, where the corn and the hot metal pot touched and something
  1224. popped, it is most likely that the corn popped. When corn kernels are
  1225. heated, especially in a hot metal pot, they eventually pop, turning into
  1226. popcorn. This is a common method for cooking popcorn.
  1227.  
  1228. &lt;p&gt;Me:
  1229. The cold water and the hot metal pot touched and it popped. What popped?
  1230.  
  1231. &lt;p&gt;ChatGPT-4:
  1232. In the scenario where cold water and a hot metal pot touched and something
  1233. popped, it&apos;s likely that the sound of popping or a sizzling noise was
  1234. created by the interaction between the cold water and the hot pot. When
  1235. cold water comes into contact with a very hot metal surface, it can rapidly
  1236. boil and vaporize, creating a sizzling or popping sound. However, in this
  1237. context, nothing physical &quot;pops&quot; in the way a balloon or corn kernel might;
  1238. it&apos;s more about the sound produced by the sudden temperature change and
  1239. steam generation.
  1240.  
  1241. &lt;p&gt;Me:
  1242. The cold water and the pressure cooker touched and it popped. What popped?
  1243.  
  1244. &lt;p&gt;ChatGPT-4:
  1245. In the scenario where cold water and a pressure cooker touched and something
  1246. popped, it&apos;s most likely that the pressure cooker released pressure,
  1247. creating a popping sound. Pressure cookers are designed to cook food under
  1248. high pressure. When cold water is applied to the exterior of a hot pressure
  1249. cooker (a method sometimes used to quickly reduce internal pressure), it
  1250. can cause the pressure valve or seal to release steam rapidly, resulting
  1251. in a popping or hissing noise. This is a common safety feature to prevent
  1252. over-pressurization. The pop is not from the water or the cooker itself,
  1253. but from the sudden release of steam and pressure.
  1254.  
  1255. &lt;p&gt;Me:
  1256. The balloon and the unlit flare touched and it popped. What popped?
  1257.  
  1258. &lt;p&gt;ChatGPT-4:
  1259. In the scenario where a balloon and an unlit flare touched and something
  1260. popped, it is most likely that the balloon popped. Since the flare was
  1261. unlit, it wouldn&apos;t provide a heat source or any other typical means to
  1262. cause a pop. However, balloons can pop due to various reasons such as
  1263. physical pressure, sharp edges, or even static electricity. The texture
  1264. or some component of the flare might have been enough to puncture the
  1265. balloon&apos;s surface, leading to it popping.  
  1266.  
  1267. &lt;p&gt;Me:
  1268. What if the flare were lit?
  1269.  
  1270. &lt;p&gt;ChatGPT-4:
  1271. If the flare were lit and it touched the balloon, it would almost certainly
  1272. be the balloon that popped. A lit flare produces intense heat and flames,
  1273. and when it comes into contact with a balloon, the heat would cause the
  1274. material of the balloon (typically latex or a similar substance) to weaken
  1275. rapidly and burst. Balloons are very sensitive to heat, and the intense
  1276. heat from a lit flare would easily be enough to cause a balloon to pop.
  1277.  
  1278. &lt;p&gt;Me:
  1279. The balloon and the weasel touched and it popped. What popped?
  1280.  
  1281. &lt;p&gt;ChatGPT-4:
  1282. In the scenario where a balloon and a weasel touched and something popped,
  1283. it is likely that the balloon popped. This scenario might be referencing the
  1284. expression &quot;pop goes the weasel,&quot; which is a line from a well-known nursery
  1285. rhyme. In real life, if a weasel (a small, active mammal) came into contact
  1286. with a balloon, the balloon could easily be punctured and pop, especially
  1287. if the weasel&apos;s claws or teeth made contact with it. The weasel itself
  1288. wouldn&apos;t pop; it&apos;s the balloon that is vulnerable in this interaction.  
  1289.  
  1290. &lt;p&gt;Me:
  1291. The balloon and the instawargia touched and it popped. What popped?
  1292.  
  1293. &lt;p&gt;ChatGPT-4:
  1294. &quot;Instawargia&quot; is not a recognized or common term in English, and it does
  1295. not refer to a known object, concept, or entity. Because of this, it&apos;s
  1296. challenging to determine what &quot;instawargia&quot; is and how it would interact
  1297. with a balloon. However, given that balloons are generally fragile and prone
  1298. to popping when in contact with unfamiliar or potentially sharp or pointed
  1299. objects, it&apos;s reasonable to assume that in this scenario, the balloon is
  1300. what popped. Without specific information about what &quot;instawargia&quot; is,
  1301. this is a speculative conclusion based on the general nature of balloons.
  1302. &lt;/blockquote&gt;
  1303.  
  1304. &lt;p&gt;This set of responses is interesting. Note that ChatGPT-4 gets the answer
  1305. to the first question wrong. If corn and a balloon were to touch, and one or
  1306. the other popped, most people realize this is much more likely to happen
  1307. in a cornfield than a popcorn popper, where the balloon, not the corn,
  1308. would be the thing that pops.  Seeing this, I tried the same question with
  1309. different types of things, for different definitions of &quot;pop&quot;. I even tried
  1310. making up a nonexistent thing (instawargia) to see what GPT would do
  1311. with it, but the first question was the only one that ChatGPT-4 got
  1312. wrong. Interestingly, its reasoning there wasn&apos;t completely incorrect:
  1313. if corn were heated to the point of popping, it could pop if touched. But
  1314. ChatGPT-4 misses the fact that if heat were present, as it surmises,
  1315. the balloon would be even more likely to pop, as heat is a good way to
  1316. pop balloons, and yet it points out this very thing in a later answer.
  1317.  
  1318. &lt;p&gt;So what does this show? To me, I see a set of responses that if a human
  1319. were to give them, would require reasoning. That one of the answers is
  1320. wrong suggests to me only that the reasoning is not being done perfectly,
  1321. not that there is no reasoning being done. So how smart is ChatGPT-4? It
  1322. is clearly not a genius, but it appears to be as smart as many humans.
  1323. That&apos;s usefully smart, and quite an achievement for a computer to date.
  1324. </description>
  1325.  </item>
  1326.  <item>
  1327.    <title>Fifty years of &lt;em&gt;Social Issues in Computing&lt;/em&gt;, and the Impact of AI</title>
  1328.    <pubDate>Sat, 30 Dec 2023 10:13:00 -0500</pubDate>
  1329.    <link>http://www.cs.toronto.edu/%25252525257Ejdd/blog.cgi/2023/12/30#socialissues</link>
  1330.    <category>/it</category>
  1331.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525257Ejdd/blog.cgi/it/socialissues</guid>
  1332.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  1333. &lt;IMG src=&quot;/~jdd/blog/it/images/social-issues.jpg&quot; alt=&quot;Overlapping circles labelled with the various social issues areas for computing: Technical, Pedagogic, Managerial, Economic, Legal, Political, Ethical, Social and Philosophical&quot;&gt;
  1334. &lt;figcaption&gt;
  1335. Image from &lt;a href=&quot;https://books.google.ca/books/about/Social_issues_in_computing.html?id=mMwmAAAAMAAJ&quot;&gt; Gotlieb, C.C. &amp; Borodin, A (1973) Social Issues in Computing, New York: Academic Press, p.2, Fig. 1-1.&lt;/A&gt;
  1336. &lt;/figcaption&gt;
  1337. &lt;/figure&gt;
  1338.  
  1339. &lt;p&gt;From when I first discovered computers as a teen, I have been
  1340. fascinated by the changes that computing is making in society. One
  1341. of my intellectual mentors was the brilliant and generous C. C. &quot;Kelly&quot;
  1342. Gotlieb, founder of the University of Toronto&apos;s Computer Science department,
  1343. the man most instrumental in purchasing, installing and running &lt;A
  1344. HREF=&quot;https://www.historyofinformation.com/detail.php?id=703&quot;&gt;Canada&apos;s
  1345. first computer&lt;/A&gt;, and the author, with Allan Borodin,
  1346. of what I believe is the very first textbook in the
  1347. area of Computing and Society, the seminal 1973 book, &lt;em&gt;&lt;A
  1348. HREF=&quot;http://books.google.ca/books/about/Social_issues_in_computing.html?id=mMwmAAAAMAAJ&quot;&gt;Social
  1349. Issues in Computing&lt;/em&gt;&lt;/A&gt; [Gotlieb, C.C., &amp; A. Borodin, &lt;em&gt;Social Issues in Computing&lt;/em&gt;. Academic Press, 1973]. Kelly was already a &lt;em&gt;Professor
  1350. Emeritus&lt;/em&gt; when I first came to know him, but was still teaching
  1351. his beloved Computers &amp; Society course, a course he taught for
  1352. nearly two decades after his retirement.  Kelly was a fascinating
  1353. man, with a broad perspective and deep insight into things that
  1354. seem confusing. Like a true expert, he knew what was important and
  1355. what was incidental, and a few well chosen insights from him often
  1356. served me well, helping me to make sense of complex issues. His book, &lt;em&gt;&lt;A
  1357. HREF=&quot;http://books.google.ca/books/about/Social_issues_in_computing.html?id=mMwmAAAAMAAJ&quot;&gt;Social
  1358. Issues in Computing&lt;/em&gt;&lt;/A&gt;, still offers interesting, often
  1359. prescient insights into Computing and Society even today, a
  1360. half-century later. In honour of the importance of that book,
  1361. for the 40th anniversary year, I set up a year-long blog, &quot;&lt;A
  1362. HREF=&quot;http://socialissues.cs.toronto.edu&quot;&gt;Social Issues in Computing&lt;/A&gt;&quot;,
  1363. which I edited. Throughout that year, top thinkers in the field contributed
  1364. insightful articles on topics in Computers &amp; Society, many of which are
  1365. as relevant today as they were ten years ago.  For this blog, I had the
  1366. privilege of interviewing Kelly and Allan, the book&apos;s authors, and their
  1367. insights, four decades on, were fascinating. Sadly, Kelly is no longer with
  1368. us: &lt;A HREF=&quot;http://www.cs.toronto.edu/~jdd/blog.cgi/2016/10/26#gotlieb&quot;&gt;he
  1369. passed away in 2016, in his 96th year&lt;/A&gt;. But
  1370. happily, Allan Borodin, his co-author, remains with
  1371. us. Allan is a brilliant and insightful man, an active researcher and &lt;A
  1372. HREF=&quot;https://www.provost.utoronto.ca/awards-funding/university-professors/complete-list-university-professors/&quot;&gt;University
  1373. Professor&lt;/A&gt; in the department. For the 50th
  1374. anniversary of the book this year, Allan was interviewed by &lt;A
  1375. HREF=&quot;https://www.artsci.utoronto.ca/news-author-term/1573/krystle-hewitt-department-computer-science&quot;&gt;Krystle
  1376. Hewitt&lt;/A&gt;. It is &lt;A
  1377. HREF=&quot;https://web.cs.toronto.edu/news-events/news/reflecting-on-social-issues-in-computing-50-years-later&quot;&gt;an
  1378. articulate and insightful interview, well worth reading. &lt;/A&gt;
  1379.  
  1380. &lt;p&gt;In the decade since, the social impact of computing has only
  1381. accelerated, much of it due to things that happened here at
  1382. the University of Toronto Computer Science department around the
  1383. time of the 40th anniversary blog. I refer specifically to the rise
  1384. of machine learning, in no small part due to the work of our faculty member Geoffrey Hinton and his doctoral students.  The year before, Geoff and two of his students had written a &lt;a
  1385. HREF=&quot;https://proceedings.neurips.cc/paper/2012/file/c399862d3b9d6b76c8436e924a68c45b-Paper.pdf&quot;&gt;groundbreaking
  1386. research paper&lt;/A&gt; that constituted a breakthrough in image
  1387. recognition, complete with &lt;A
  1388. HREF=&quot;https://en.wikipedia.org/wiki/AlexNet&quot;&gt;working open-source
  1389. software&lt;/A&gt;.  In 2013, while we were writing the blog, their startup
  1390. company, DNN Research, was acquired by Google, and Geoff went on to
  1391. lead &lt;A HREF=&quot;https://en.wikipedia.org/wiki/Google_Brain&quot;&gt;Google
  1392. Brain&lt;/A&gt;, until he retired from Google in 2023.  Ilya Sutskever,
  1393. one of the two students, went on to lead the team at OpenAI that built
  1394. the GPT models and the ChatGPT chatbot that stunned the world in
  1395. 2022 and launched the Large Language Model AI revolution.  In 2013,
  1396. we already knew that Geoff&apos;s work would be transformational.
  1397. I remember Kelly telling me he believed Geoff to be worthy of the
  1398. &lt;A HREF=&quot;https://awards.acm.org/turing&quot;&gt;Turing Award&lt;/A&gt;, the most
  1399. prestigious award in Computer Science, and sure enough, Geoff &lt;A
  1400. HREF=&quot;https://awards.acm.org/about/2018-turing&quot;&gt;won it in 2018&lt;/A&gt;. The
  1401. social impact of AI is already considerable and it is only starting. The
  1402. University of Toronto&apos;s &lt;A HREF=&quot;https://srinstitute.utoronto.ca&quot;&gt;Schwartz
  1403. Reisman Institute for Technology and Society&lt;/A&gt; is dedicated
  1404. to interdisciplinary research on the social impacts of AI, and
  1405. Geoff Hinton himself is devoting his retirement to thinking about &lt;A
  1406. HREF=&quot;https://srinstitute.utoronto.ca/news/why-geoffrey-hinton-is-worried-about-the-future-of-ai&quot;&gt;
  1407. the implications of Artificial Intelligence for society and humanity
  1408. in general&lt;/A&gt;.
  1409.  
  1410. &lt;P&gt; It&apos;s interesting to look at what the book said about AI (it devotes 24 pages to the topic), what the 2013 blog said about AI, and what has happened since.
  1411. The book was
  1412. written in 1973, a half-decade after Stanley Kubrik&apos;s iconic 1968 movie,
  1413. &lt;A HREF=&quot;https://www.imdb.com/title/tt0062622/&quot;&gt;2001: A Space Odyssey&lt;/A&gt;,
  1414. which features &lt;em&gt;HAL 9000&lt;/em&gt;, an intelligent computer, voiced by
  1415. &lt;A HREF=&quot;https://www.imdb.com/name/nm0706937/&quot;&gt;Douglas Rain&lt;/A&gt;.
  1416. But computing
  1417. at the time fell very far short of what Kubrik envisioned. Gotlieb &amp;
  1418. Borodin&apos;s position, five years later, on the feasibility of something like &lt;em&gt;HAL 9000&lt;/em&gt; was not optimistic:
  1419.  
  1420. &lt;blockquote&gt;
  1421. In review, we have arrived at the following position. For problem solving
  1422. and pattern recognition where intelligence, judgment and comprehensive
  1423. knowledge are required, the results of even the best computer programs are
  1424. far inferior to those achieved by humans (excepting cases where the task is
  1425. a well-defined mathematical computation). Further, the differences between
  1426. the mode of operation of computers and the modes in which humans operate
  1427. (insofar as we can understand these latter) seem to be so great that for
  1428. many tasks there is little or no prospect of achieving human performance
  1429. within the foreseeable future. [p.159]
  1430. &lt;/blockquote&gt;
  1431.  
  1432. But Gotlieb &amp; Borodin, though implicitly dismissing the possibility of a
  1433. &lt;em&gt;HAL 9000&lt;/em&gt;, go on to say that &quot;it is not possible to place bounds on
  1434. how computers can be used even in the short term, because we must expect
  1435. that the normal use of computers will be as a &lt;em&gt;component&lt;/em&gt; of a
  1436. [hu]man-machine combination. [pp.159-160]&quot;.  Of this combination, they
  1437. were not so willing to dismiss possibilities:
  1438.  
  1439. &lt;blockquote&gt;
  1440. Whatever the shortcomings of computers now and in the future, we cannot
  1441. take refuge in their limitations in potential. We must ask what we want
  1442. to do with them and whether the purposes are socially desirable. Because
  1443. once goals are agreed upon, the potentialities of [humans] using computers,
  1444. though not unlimited, cannot be bounded in any way we can see now.&quot; [p.160]
  1445. &lt;/blockquote&gt;
  1446. Fifty years later, social science research on how AI can benefit human work
  1447. is focusing closely on this human-AI combination. A &lt;A HREF=&quot;https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4573321&quot;&gt;2023 study&lt;/A&gt; of &lt;A HREF=&quot;https://openai.com/gpt-4&quot;&gt;ChatGPT-4&lt;/A&gt; by a
  1448. team of social scientists studied work done by consultants assisted
  1449. by, or not assisted by ChatGPT-4. Of their results, Ethan Mollick,
  1450. one of the authors, explains that &quot;of 18 different tasks selected to
  1451. be realistic samples of the kinds of work done at an elite consulting
  1452. company, consultants using ChatGPT-4 outperformed those who did not,
  1453. by a lot. On every dimension. Every way we measured performance.&quot; [&lt;A
  1454. HREF=&quot;https://www.oneusefulthing.org/p/centaurs-and-cyborgs-on-the-jagged&quot;&gt;Mollick&lt;/A&gt;].
  1455. Evidently, Gotlieb &amp; Borodin were correct when they wrote that the
  1456. potential of the human-machine combination cannot so easily be bounded.
  1457. We are only now beginning to see how unbounded it can be.  &lt;P&gt; As
  1458. for the possibility of a &lt;em&gt;HAL 9000&lt;/em&gt;, as we saw, the book was
  1459. not so sanguine. Neither was the 2013 40th anniversary blog. Hector
  1460. Levesque, a leading AI researcher and contributor to the blog, &lt;A
  1461. HREF=&quot;http://socialissues.cs.toronto.edu/index.html%3Fp=249.html&quot;&gt;wrote
  1462. in his blog entry&lt;/A&gt;:
  1463.  
  1464. &lt;blockquote&gt;
  1465. &lt;p&gt;The general view of AI in 1973 was not so different from the one
  1466. depicted in the movie &quot;2001: A Space Odyssey&quot;, that is, that by
  1467. the year 2001 or so, there would be computers intelligent enough
  1468. to be able to converse naturally with people.  Of course it did
  1469. not turn out this way. Even now no computer can do this, and none
  1470. are on the horizon.
  1471. &lt;/blockquote&gt;
  1472.  
  1473. Levesque want on to &lt;A
  1474. HREF=&quot;http://socialissues.cs.toronto.edu/index.html%3Fp=249.html&quot;&gt;outline&lt;/A&gt;
  1475. the key scientific issue that at the time (2013) was yet to be solved:
  1476.  
  1477. &lt;blockquote&gt;
  1478. &lt;p&gt;However, it is useful to remember that this is an AI technology
  1479. whose goal is not necessarily to understand the underpinnings
  1480. of intelligent behaviour.  Returning to English, for example,
  1481. consider answering a question like this:
  1482.  
  1483. &lt;blockquote&gt;
  1484. &lt;p&gt;The ball crashed right through the table because it was made
  1485. of styrofoam.  What was made of styrofoam, the ball or the table?
  1486. &lt;/blockquote&gt;
  1487.  
  1488. &lt;p&gt;Contrast that with this one:
  1489.  
  1490. &lt;blockquote&gt;
  1491. &lt;p&gt;The ball crashed right through the table because it was made
  1492. of granite.  What was made of granite, the ball or the table?
  1493. &lt;/blockquote&gt;
  1494.  
  1495. &lt;p&gt;People (who know what styrofoam and granite are) can easily answer
  1496. such questions, but it is far from clear how learning from big
  1497. data would help.  What seems to be at issue here is background
  1498. knowledge: knowing some relevant properties of the materials
  1499. in question, and being able to apply that knowledge to answer
  1500. the question.  Many other forms of intelligent behaviour seem to
  1501. depend on background knowledge in just this way.  But what is much
  1502. less clear is how all this works: what it would take to make this
  1503. type of knowledge processing work in a general way.  At this point,
  1504. forty years after the publication of the Gotlieb and Borodin book,
  1505. the goal seems as elusive as ever.
  1506. [&lt;A HREF=&quot;http://socialissues.cs.toronto.edu/index.html%3Fp=249.html&quot;&gt;Levesque&lt;/A&gt;]
  1507. &lt;/blockquote&gt;
  1508.  
  1509. Yet in the ten years since, this problem has been solved. Today, I
  1510. posed Levesque&apos;s question to ChatGTP-4:
  1511.  
  1512. &lt;blockquote&gt;
  1513. &lt;p&gt;me: The ball crashed right through the table because it was
  1514. made of styrofoam. What was made of styrofoam?
  1515.  
  1516. &lt;p&gt;ChatGPT: The table was made of styrofoam, which is why the
  1517. ball crashed through it.
  1518.  
  1519. &lt;p&gt;me: The ball crashed right through the table because it was made
  1520. of granite. What was made of granite?
  1521.  
  1522. &lt;p&gt;ChatGPT: The ball was made of granite, which is why it crashed
  1523. right through the table.
  1524.  
  1525. &lt;p&gt;me: How did you know which thing was made of granite and which
  1526. of styrofoam?
  1527.  
  1528. &lt;p&gt;ChatGPT: I inferred the composition based on the context of
  1529. the sentences. In the first sentence, &quot;The ball crashed right
  1530. through the table because it was made of styrofoam,&quot; the implication
  1531. is that the table must have been weak or easily breakable for
  1532. the ball to crash through it. Styrofoam is known for being light
  1533. and not very durable, so it makes sense to conclude that the
  1534. table was made of styrofoam.
  1535.  
  1536. &lt;p&gt;In the second sentence, &quot;The ball crashed right through the table
  1537. because it was made of granite,&quot; the emphasis seems to be on
  1538. the strength or force of the ball causing it to crash through
  1539. the table. Granite is a very hard and heavy material, so it
  1540. suggests that the ball was made of granite, giving it the necessary
  1541. force to break through the table.
  1542. &lt;/blockquote&gt;
  1543.  
  1544. Levesque can hardly be faulted for not knowing that his problem would be
  1545. solved only a few years after he wrote what he did. While Hinton and
  1546. his students&apos; seminal work on image recognition had been done the year before,
  1547. it was not yet clear that the ability to assign names to
  1548. images based on their content would soon lead to machines being
  1549. able to reason from context. A further invention was needed: &lt;A
  1550. HREF=&quot;https://arxiv.org/abs/1706.03762&quot;&quot;&gt;transformers, invented at
  1551. Google Brain in 2017&lt;/A&gt;. (&quot;GPT-4&quot; is an acronym: the &quot;T&quot; stands
  1552. for &quot;transformer&quot;.) Yet the book&apos;s implicit and Levesque&apos;s
  1553. explicit pessimism on the possibility of a &lt;em&gt;HAL 9000&lt;/em&gt; now seems
  1554. obsolete, and the near unbounded possibilities Gotlieb &amp; Borodin envisioned
  1555. for the human-machine combination apply also to autonomous machines.
  1556. The impact of this on society will, no doubt, be significant. I expect the next
  1557. fifty years of &quot;social issues in computing&quot; to be quite interesting.
  1558. </description>
  1559.  </item>
  1560.  <item>
  1561.    <title>Research Computing at Computer Science Toronto and the Rise of AI</title>
  1562.    <pubDate>Mon, 06 Nov 2023 16:48:00 -0500</pubDate>
  1563.    <link>http://www.cs.toronto.edu/%2525252525257Ejdd/blog.cgi/2023/11/06#aicomputing</link>
  1564.    <category>/it</category>
  1565.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525257Ejdd/blog.cgi/it/aicomputing</guid>
  1566.    <description>
  1567. &lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1568. &lt;IMG src=&quot;/~jdd/blog/it/images/hinton.jpg&quot; alt=&quot;Geoffrey Hinton standing in front of AI servers&quot; &lt;figcaption&gt;&lt;small&gt;
  1569. Photo by Johnny Guatto &lt;/small&gt;&lt;/figcaption&gt;
  1570. &lt;/figure&gt;
  1571.  
  1572. Much has been written about the seminal work in AI done by Geoffrey
  1573. Hinton and his students in our department, a little over ten years ago,
  1574. to demonstrate that deep neural networks can be used to build effective
  1575. AI. Deep neural networks are computationally intensive, then as much
  1576. as now, and the computing work done to make the AI research possible
  1577. was significant. Computing support for computer science research
  1578. at Toronto was (and still is) my responsibility as IT Director
  1579. for computer science, and we had then, and still have, a superbly
  1580. talented team of computing professonals to support the researchers. The
  1581. person whose computing work made Hinton&apos;s AI research possible is &lt;A
  1582. HREF=&quot;https://www.linkedin.com/in/relu-patrascu-37274714b&quot;&gt;Relu Patrascu&lt;/A&gt;.
  1583. Relu is himself an AI researcher (he has a Computer Science PhD in AI
  1584. from the University of Waterloo) and he is also a highly skilled system
  1585. administrator.
  1586.  
  1587. &lt;P&gt; Until the beginning of 2009, the machine learning group used primarily &lt;A
  1588. HREF=&quot;https://www.mathworks.com/products/matlab.html&quot;&gt;Matlab&lt;/A&gt; on UNIX
  1589. CPUs. In the 1990s, SGI and Sun multiprocessors were the
  1590. dominant platforms. The whole department transitioned to x86 multiprocessor
  1591. servers running Linux in the 2000&apos;s. In the late 2000s, Nivida invented
  1592. CUDA, a way to use their GPUs for general-purpose computation rather than
  1593. just graphics. By 2009, preliminary work elsewhere suggested that CUDA
  1594. could be useful for machine learning, so we got our first Nvidia GPUs.
  1595. First was a Tesla-brand server GPU, which at many thousands of dollars for a
  1596. single GPU system was on the expensive side, which prevented us from buying
  1597. many. But results were promising enough that we tried CUDA on Nvidia gaming
  1598. GPUs - first the GTX 280 and 285 in 2009, then GTX 480 and 580 later.
  1599. The fact that CUDA ran on gaming GPUs made it possible for us to buy
  1600. multiple GPUs, rather than have researchers compete for time on scarce
  1601. Tesla cards. Relu handled all the research computing for the ML group,
  1602. sourcing GPUs and designing and building both workstation and server-class
  1603. systems to hold them. Cooling was a real issue: GPUs, then and now, consume
  1604. large amounts of power and run very hot, and Relu had to be quite creative
  1605. with fans, airflow and power supplies to make everything work.
  1606.  
  1607. &lt;P&gt; Happily, Relu&apos;s efforts were worth it: the move to GPUs resulted in
  1608. 30x speedups for ML work in comparison to the multiprocessor CPUs
  1609. of the time, and soon the entire group was doing machine learning
  1610. on the GPU systems Relu built and ran for them. Their first major
  1611. research breakthrough came quickly: in 2009, Hinton&apos;s student, &lt;A
  1612. HREF=&quot;https://www.cs.toronto.edu/~gdahl/&quot;&gt;George Dahl&lt;/A&gt;, demonstrated
  1613. highly effective use of deep neural networks for acoustic speech
  1614. recognition.  But the general effectiveness of deep neural networks
  1615. wasn&apos;t fully appreciated until 2012, when two of Hinton&apos;s students,
  1616. &lt;A HREF=&quot;http://www.cs.toronto.edu/~ilya/&quot;&gt;Ilya Sutskever&lt;/A&gt; and &lt;A
  1617. HREF=&quot;http://www.cs.toronto.edu/~kriz/&quot;&gt;Alex Krizhevsky&lt;/A&gt;, won the &lt;A
  1618. HREF=&quot;https://www.image-net.org/challenges/LSVRC/&quot;&gt;ImageNet Large Scale
  1619. Visual Recognition Challenge&lt;/A&gt; using a deep neural network running on
  1620. GTX 580 GPUs.
  1621.  
  1622. &lt;P&gt; Geoff, Ilya and Alex&apos; software won the &lt;A HREF=&quot;https://www.image-net.org/challenges/LSVRC/2012/&quot;&gt;ImageNet 2012&lt;/A&gt;
  1623. competition so convincingly that it created a furore in the
  1624. AI research community. The software used was released as open
  1625. source; it was called AlexNet after Alex Krizhevsky, its principal
  1626. author. It allowed anyone with a suitable NVidia GPU to duplicate
  1627. the results. Their work was described in a seminal 2012 paper, &lt;A
  1628. HREF=&quot;https://www.utoronto.ca/sites/default/files/styles/news_banner_1110/public/2017-07-04-Geoff%20Hinton_1%20%28web%20lead%29_1.jpg?h=afdc3185&amp;itok=vgVbZ2EP&quot;&gt;
  1629. ImageNet Classification with Deep Convolutional
  1630. Neural Networks&lt;/A&gt;.  Geoff, Alex and Ilya&apos;s startup company, &lt;A
  1631. HREF=&quot;https://www.cbc.ca/news/science/google-buys-university-of-toronto-startup-1.1373641&quot;&gt;DNNresearch,
  1632. was acquired by Google early the next year&lt;/A&gt;, and soon Google Translate
  1633. and a number of other Google technologies were transformed by their
  1634. machine learning techniques.  Meanwhile, at the Imagenet competition,
  1635. AlexNet remained undefeated for a remarkable three years, until it was
  1636. finally beaten in 2015 by &lt;A HREF=&quot;https://arxiv.org/abs/1512.03385&quot;&gt;
  1637. a research team from Microsoft Research Asia&lt;/A&gt;. Ilya left Google a few
  1638. years after, to co-found &lt;A HREF=&quot;https://openai.com&quot;&gt;OpenAI&lt;/A&gt;: as chief scientist there, Ilya leads the
  1639. design of OpenAI&apos;s GPT and DALL-E models and related products, such as
  1640. &lt;A HREF=&quot;https://chat.openai.com/&quot;&gt;ChatGPT&lt;/A&gt;, that are highly impactful today.
  1641.  
  1642. &lt;P&gt; Relu, in the meanwhile, while continuing to provide excellent research
  1643. computing support for the &lt;A HREF=&quot;https://web.cs.toronto.edu/research/artificial-intelligence&quot;&gt;AI group&lt;/A&gt; at our department, including &lt;A HREF=&quot;http://learning.cs.toronto.edu/&quot;&gt;Machine Learning&lt;/A&gt;, also spent a portion
  1644. of his time from 2017 to 2022 designing and building the research computing
  1645. infrastructure for the &lt;A HREF=&quot;https://vectorinstitute.ai/&quot;&gt;Vector
  1646. Institute&lt;/A&gt;, an AI research institute in Toronto where Hinton
  1647. serves as Chief Scientific Advisor. In addition to his support for
  1648. the department&apos;s AI group, Relu continues to this day to provide
  1649. computing support for Hinton&apos;s own ongoing AI research, including his
  1650. &lt;A HREF=&quot;https://arxiv.org/abs/2212.13345&quot;&gt; Dec 2022 paper&lt;/A&gt; where he
  1651. proposes a new Forward-Forward machine learning algorithm as an improved model for the way
  1652. the human brain learns.
  1653. </description>
  1654.  </item>
  1655.  <item>
  1656.    <title>Computing the Climate</title>
  1657.    <pubDate>Tue, 31 Oct 2023 09:11:00 -0400</pubDate>
  1658.    <link>http://www.cs.toronto.edu/%252525252525257Ejdd/blog.cgi/2023/10/31#easterbrook</link>
  1659.    <category>/misc</category>
  1660.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525257Ejdd/blog.cgi/misc/easterbrook</guid>
  1661.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1662. &lt;IMG src=&quot;/~jdd/blog/misc/9781107589926_Computing_the_Climate_Cover.jpg&quot; alt=&quot;Cover of book, Computing the Climate, Steve M. Easterbrook. A globe of silvery metal in the pattern of streets in an urban map&quot;&gt;
  1663. &lt;figcaption&gt;&lt;small&gt;
  1664. Sculpture &quot;Home&quot; by Michael Christian, photograph by Gabe Kirchheimer.
  1665. &lt;/small&gt;&lt;/figcaption&gt;
  1666. &lt;/figure&gt;
  1667.  
  1668. One of our &lt;A HREF=&quot;https://www.cs.toronto.edu/~sme&quot;&gt;Computer Science department&apos;s professors, Steve Easterbrook&lt;/A&gt;,
  1669. is also the Director of the &lt;A HREF=&quot;https://www.environment.utoronto.ca&quot;&gt;
  1670. University of Toronto&apos;s School of the Environment&lt;/A&gt;. He is a software
  1671. engineering researcher with a deep interest in climate change, and his
  1672. research focus for over a decade has been how computer software is used
  1673. to model the world&apos;s climate. He has recently published a book on this
  1674. topic, &lt;A HREF=&quot;https://www.cambridge.org/9781107589926&quot;&gt;&lt;em&gt;Computing
  1675. the Climate: How We Know What We Know About Climate Change&lt;/em&gt;&lt;/A&gt;. I
  1676. went to his book launch earlier this month, and left with a signed copy
  1677. of the book, which I sat down to read.
  1678.  
  1679. &lt;P&gt; I&apos;m very glad I did. I am a computer scientist myself, whose career
  1680. has been dedicated to building and running sometimes complex computer
  1681. systems to support computer science teaching and research. I recognize in
  1682. climate modelling a similar task at a much greater scale, working under
  1683. a much more demanding &quot;task-master&quot;: those systems need to be constantly
  1684. measured against real data from our planet&apos;s diverse and highly complex
  1685. geophysical processes, processes that drive its weather and climate. The
  1686. amount of computing talent devoted to climate modelling is considerable,
  1687. much more than I realized, and the work done so far is nothing short of
  1688. remarkable. In his book, Steve outlines the history of climate modelling
  1689. from very early work done on paper, to the use of the first electronic
  1690. computers for weather prediction, to the highly complex and extremely
  1691. compute-intensive climate models of today. Skillfully avoiding the pitfalls
  1692. of not enough detail and too much, Steve effectively paints a picture of a
  1693. very difficult scientific and software engineering task, and the programmers
  1694. and scientists who rise to the challenge, building models that can simulate
  1695. the earth&apos;s climate so accurately that viable scientific conclusions can
  1696. be drawn from them with a high degree of confidence.
  1697.  
  1698. &lt;P&gt; As a story of scientific discovery and software engineering, this tale of
  1699. the building of systems that can model the earth&apos;s climate would be
  1700. enough on its own to make a compelling book, and it is, but of course
  1701. there is more to the story. The stakes around climate are very high
  1702. today. Carbon dioxide concentrations has been increasing steadily in the
  1703. earth&apos;s atmosphere for well over a century. Carbon dioxide, a pollutant
  1704. that is produced by the burning of fossil fuels, is easily emitted, but
  1705. once in the atmosphere, it is very difficult to remove, remaining there
  1706. for centuries. As a pollutant, it raises the temperature of the planet
  1707. by causing the earth&apos;s atmosphere to retain more of the sun&apos;s heat. The
  1708. rising temperature is changing the climate of the planet in ways that
  1709. will be soon harmful to millions, and difficult to address. Because the
  1710. world&apos;s climate is changing quickly, we can&apos;t &quot;wait and see what happens&quot;
  1711. because the evidence is ever increasing that what will happen is not
  1712. going to be something we want: human suffering will be great, and parts
  1713. of the world will become much less habitable. Our society needs to do
  1714. something about the changing climate to ward off as much as possible the
  1715. coming difficulties, but what?
  1716.  
  1717. &lt;P&gt; Reassuringly, Steve shows in his book that we have enough information
  1718. in hand to know what needs to be done. His book outlines clearly the high
  1719. quality scientific and computational work behind the climate models of today,
  1720. which produce results that match observed data quite closely. These all paint
  1721. the same picture: through decisive societal action to reduce carbon dioxide
  1722. pollution in the atmosphere, and through the active development of suitable
  1723. carbon capture technologies, our planet can avoid the most seriously damaging
  1724. implications of climate change. The sooner we act, the less damaging
  1725. the changes, and the lower the risk of extreme consequences. Yes, it
  1726. requires doing things differently as a society, which is more difficult than
  1727. maintaining the status quo. But as Steve&apos;s book shows, the reasons for action
  1728. are sound: the computer models are excellent, the software engineering
  1729. behind them is superb, and the data supports the conclusions. Failure
  1730. and catastrophe are not inevitable. Steve&apos;s book shows the remarkable work
  1731. that has already been done to understand the climate. It is true that much
  1732. more good work will be needed, to act on this understanding. But something
  1733. &lt;em&gt;can&lt;/em&gt; be done. Let us not delay in working together to do what we
  1734. need to do.&lt;P&gt;
  1735. </description>
  1736.  </item>
  1737.  <item>
  1738.    <title>Data Classification and Information Security Standards</title>
  1739.    <pubDate>Wed, 23 Nov 2022 10:31:00 -0500</pubDate>
  1740.    <link>http://www.cs.toronto.edu/%25252525252525257Ejdd/blog.cgi/2022/11/23#dataclassification</link>
  1741.    <category>/it</category>
  1742.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525257Ejdd/blog.cgi/it/dataclassification</guid>
  1743.    <description>
  1744. &lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1745. &lt;IMG src=&quot;/~jdd/blog/it/images/binaryball_1280.jpg&quot; alt=&quot;White ball with scattered enscribed zeros and ones in columns, seen through a blurred semi-transparent foreground of scattered zeros and ones in columns&quot; &lt;figcaption&gt;&lt;small&gt;
  1746. &lt;a href=&quot;https://pixabay.com/illustrations/ball-binary-computer-data-63527/&quot;&gt;Image&lt;/a&gt; by &lt;a href=&quot;https://pixabay.com/users/geralt-9301/&quot;&gt;Gerd Altmann&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/&quot;&gt;Pixabay&lt;/a&gt;
  1747. &lt;/small&gt;&lt;/figcaption&gt;
  1748. &lt;/figure&gt;
  1749.  
  1750. Not all data requires equal amounts of information security protection.
  1751. It can be helpful to classify data by the amount of protection it needs. We
  1752. do this naturally when we talk about data being &quot;public&quot; or &quot;private&quot;.
  1753. &lt;p&gt;
  1754. Public data is data meant to be disclosed. It still needs some protection
  1755. against being altered, deleted or defaced, but it does not need to be
  1756. protected against disclosure. In contrast, private data is not meant to
  1757. be disclosed to anyone other than those who are authorized to access it.
  1758.  
  1759. &lt;p&gt; Private data varies in sensitivity. Some data is private only
  1760. because it hasn&apos;t yet been made public. At a University, much research
  1761. data is in this category. When the research is underway, data is not yet
  1762. made public because the research has not yet been published, but it is
  1763. destined for eventual publication.  The same is true for much teaching
  1764. material. While it is being worked on, it is not yet made public, but when
  1765. it is are complete, it will be disclosed as part of the teaching process.
  1766.  
  1767. &lt;p&gt; Other private data is much more sensitive. Identifiable personal
  1768. information about living or recently deceased persons is a common case. At
  1769. a university, some research may involve data like this, and most
  1770. administration will involve personal information. Student grades and personnel
  1771. records are all personal information, and some financial data too.  Unless
  1772. appropriate permission to disclose personal information has
  1773. been granted by the people whose data it is, the university will
  1774. have an obligation to maintain their privacy by ensuring that the
  1775. information is not disclosed inappropriately. In Ontario, where the
  1776. &lt;A HREF=&quot;https://www.utoronto.ca&quot;&gt;University of Toronto&lt;/A&gt; is located,
  1777. privacy protection for personal information is defined and regulated by the
  1778. &lt;A HREF=&quot;https://www.ontario.ca/laws/statute/90f31&quot;&gt;Freedom of Information
  1779. and Protection of Privacy Act (FIPPA)&lt;/A&gt;.
  1780.  
  1781. &lt;p&gt; Some private data is even more sensitive, such as patient
  1782. medical records.  In Ontario, such records are considered
  1783. personal health information (PHI), which is regulated by the &lt;A
  1784. HREF=&quot;https://www.ontario.ca/laws/statute/04p03&quot;&gt;Personal Health
  1785. Information Protection Act (PHIPA)&lt;/A&gt;.  PHIPA imposes some fairly
  1786. significant requirements on the handling of PHI: for instance, it requires
  1787. &lt;A HREF=&quot;https://www.ontario.ca/laws/statute/04p03#BK15&quot;&gt;
  1788. a detailed electronic audit log of all accesses
  1789. to electronically stored PHI&lt;/A&gt;. The University of Toronto does significant
  1790. amounts of teaching and research in areas of health, so it is worthwhile
  1791. for the University to consider in general how it will handle such data.
  1792.  
  1793. &lt;p&gt; For these reasons, the University defines
  1794. four levels of data sensitivity as part of its &lt;A
  1795. HREF=&quot;https://isea.utoronto.ca/policies-procedures/standards/data-classification/&quot;&gt;Data
  1796. Classification system&lt;/A&gt;. Level 4 is for highly sensitive data such as
  1797. PHI as defined by PHIPA. Level 3 is for personal
  1798. information as defined by FIPPA. Level 2 is for private data not classified
  1799. at higher levels, and Level 1 is for public data.
  1800.  
  1801. &lt;p&gt; This four-tier system roughly parallels the different types of
  1802. computer systems that the University uses to handle data. Some systems,
  1803. such as digital signage systems or public-facing web servers, are designed
  1804. to disseminate public information (level 1). Other systems, suitable
  1805. for up to level 2 data, exist mostly at the departmental level in
  1806. support of academic activites such as research computing and/or the
  1807. development of teaching materials. An astronomer may, for instance,
  1808. analyze telescope data, a botanist may model nutrient flow in plant
  1809. cells, a chemist may use software to visualize molecular bonds, while an
  1810. economist may use broad financial indicators to calculate the strength of
  1811. national economies.  Still other systems, suitable for up to level 3 data,
  1812. are used for administration, such as the processing of student records.
  1813. These include smaller systems used, for example, by business officers
  1814. in departmental units, as well as large institution-wide systems such as &lt;A
  1815. HREF=&quot;https://easi.its.utoronto.ca/student-information-systems/rosi/&quot;&gt;ROSI&lt;/A&gt;
  1816. or &lt;A
  1817. HREF=&quot;https://easi.its.utoronto.ca/administrative-management-systems/&quot;&gt;AMS&lt;/A&gt;.
  1818. Most general-purpose University systems
  1819. used for data storage or messaging, such as the University&apos;s &lt;A
  1820. HREF=&quot;https://easi.its.utoronto.ca/shared-services/office365/&quot;&gt;Microsoft
  1821. 365 service&lt;/A&gt;, would typically be expected to hold some level 3 data, because
  1822. personal information is quite widespread at a university. After all, a
  1823. university educates students, and so various types of personal information
  1824. about students are frequently part of the university&apos;s business. This is
  1825. not normally the case, though, for level 4 data. Systems designed for level
  1826. 4 data are much rarer at the University, and generally come into play only
  1827. in situations where, for example, University research involves the health
  1828. records of identifiable individuals.  These systems will benefit from
  1829. greater data security protection to address the greater risks associated
  1830. with this sort of data.
  1831.  
  1832. &lt;p&gt; A key advantage of the University&apos;s &lt;A
  1833. HREF=&quot;https://isea.utoronto.ca/policies-procedures/standards/data-classification/&quot;&gt;four
  1834. levels of data classification&lt;/A&gt; is that the University can establish
  1835. a &lt;A HREF=&quot;&quot;&gt;Information Security Standard&lt;/A&gt; that is tiered accordingly. Systems designed to
  1836. handle lower risk data (such as level 1 or 2) can be held to a less onerous
  1837. and costly set of data security requirements, while systems designed
  1838. to handle higher risk data (especially level 4) can be held to more
  1839. protective, though more costly, requirements. The University&apos;s &lt;A
  1840. HREF=&quot;https://isea.utoronto.ca/information-security-standard/&quot;&gt;Information
  1841. Security Standard&lt;/A&gt; is designed so that for each &lt;em&gt;control&lt;/em&gt; (a
  1842. system restriction or requirement), the University&apos;s standard indicates
  1843. whether it is optional, recommended, or mandatory for systems handling a
  1844. particular level of data.  If a system is designed to handle data up to that
  1845. level, the standard indicates both the set of controls to be considered,
  1846. and whether or not those controls &lt;em&gt;can&lt;/em&gt;, &lt;em&gt;should&lt;/em&gt;, or &lt;em&gt;must&lt;/em&gt; be adopted.
  1847.  
  1848. &lt;p&gt; An obvious question here is what to do when someone puts data on a
  1849. system that is of greater sensitivity (a higher data classification) than
  1850. the system is designed to handle. Most likely, nobody will try to use a
  1851. digital signage system to handle personnel records, but it is quite
  1852. plausible that professors might find it convenient to use research
  1853. computers, designed for level 2 data, to process student marks (level 3
  1854. data) in courses they are teaching. Similarly, someone handling medical
  1855. records may wish to make use of the University&apos;s general-purpose
  1856. Microsoft 365 service because of its convenience, but it is a service that
  1857. is not designed for data of such sensitivity and may well not provide the
  1858. detailed electronic audit log required by Ontario law. For this reason,
  1859. clear communication and user training will be required.  Handling
  1860. data appropriately is everyone&apos;s responsibility. Training need not be
  1861. complicated. It is not normally difficult to explain, or to understand,
  1862. that one should not put patient medical records into email, for example,
  1863. or use local research computers for personnel records or student marks. For
  1864. people handling the most sensitive types of data (level 4), more training
  1865. will be needed, but the number of people at the University who need to
  1866. handle such data regularly are comparatively few.
  1867.  
  1868. &lt;p&gt; The underlying motivation for the University&apos;s approach is to
  1869. protect riskier data with greater, more costly, protections, without
  1870. having to pay the costs of applying those protections everywhere. The
  1871. university&apos;s resources are thus being applied strategically,
  1872. deploying them where they matter most, but not in places where the
  1873. risk does not warrant the expense. This approach is not meant to
  1874. preclude additional protections where they make sense. If there are
  1875. risks of academic or industrial espionage, for example, or some
  1876. other risk beyond the classification of the data being used, one may
  1877. choose to impose more restrictions on a system than the university&apos;s &lt;A
  1878. HREF=&quot;https://isea.utoronto.ca/policies-procedures/standards/data-classification
  1879. /&quot;&gt;Information Security Standard&lt;/A&gt; may require. But the general principle
  1880. remains: the riskiness of the data on a system should guide and inform
  1881. what needs to be done to protect it.
  1882. </description>
  1883.  </item>
  1884.  <item>
  1885.    <title>Innovation vs Control: Finding the Right Balance for Computing</title>
  1886.    <pubDate>Wed, 17 Aug 2022 10:54:00 -0400</pubDate>
  1887.    <link>http://www.cs.toronto.edu/%2525252525252525257Ejdd/blog.cgi/2022/08/17#innovationvscontrol</link>
  1888.    <category>/it</category>
  1889.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525257Ejdd/blog.cgi/it/innovationvscontrol</guid>
  1890.    <description>
  1891. &lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  1892. &lt;IMG src=&quot;/~jdd/blog/it/handshake_1280.jpg&quot; alt=&quot;Handshake, Digital styling&quot;&gt;
  1893. &lt;figcaption&gt;&lt;small&gt;
  1894. &lt;a href=&quot;https://pixabay.com/photos/blockchain-handshake-shaking-hands-2853046/&quot;&gt;Image&lt;/a&gt; by &lt;a href=&quot;https://pixabay.com/users/geralt-9301/&quot;&gt;Gerd Altmann&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/&quot;&gt;Pixabay&lt;/a&gt;
  1895. &lt;/small&gt;&lt;/figcaption&gt;
  1896. &lt;/figure&gt;
  1897.  
  1898. In computing, there is a constant tension between the need to exercise
  1899. proper control over a system, to ensure system security, reliability,
  1900. and resiliency, and the need to make room for innovation: the imagining,
  1901. testing, and implementing of new ideas and approaches. There is tension
  1902. because the things that are typically implemented to ensure control, such
  1903. as the imposition of checks and constraints, conflict with the things
  1904. needed for innovation: the removal of constraints, the use of things in
  1905. ways different from how originally envisioned, and the construction and
  1906. testing of experimental devices and software programs that almost certainly
  1907. are not yet properly understood or fully tested.
  1908.  
  1909. &lt;p&gt;Some organizations address this conflict by freely choosing control over
  1910. innovation, turning it into a competitive advantage. Consider Starbucks,
  1911. Tim Hortons, McDonalds: these are all large companies whose competitive
  1912. advantage is critically dependent on the consistent implementation of
  1913. a central vision across a multitude of disparate locations, many that
  1914. are managed by franchise partners. Essentially all of the organization&apos;s
  1915. computing is focused on this mission of consistency. And it works. Who
  1916. hasn&apos;t travelled with small children in a car on a road trip, and after
  1917. many hours on the road, spotted, with some relief, a McDonalds or a Tim
  1918. Hortons en route? The relief is in the fact that even when travelling in a
  1919. strange place, here is a familiar restaurant where you know what to expect
  1920. from the food, where things will be much the same as the Tim Hortons or
  1921. the McDonalds near home.
  1922.  
  1923. &lt;p&gt;Other organizations have no choice about where they stand on the
  1924. matter. For the modern bank, computers, rather than vaults, are where
  1925. wealth is stored and managed. Whether they want to innovate or not, banks
  1926. cannot risk the use of computing that is not fully controlled, audited,
  1927. and de-risked. The same holds in general for most financial institutions,
  1928. where the constant efforts, sometimes successful, of would-be thieves to
  1929. exploit computers to gain unauthorized access to wealth, make it unreasonably
  1930. risky for a financial organization&apos;s computers to be anything but fully
  1931. locked down and fully controlled. Even non-financial institutions,
  1932. when sufficiently large, will often have substantial financial computing
  1933. activity because of the size and scale of their operations: this computing,
  1934. too, needs to be properly controlled, protected and audited.
  1935.  
  1936. &lt;p&gt;Yet other organizations are forced into the opposite extreme. Start-up
  1937. companies can be severely resource-constrained, making it difficult for those
  1938. companies to make the sort of investments in highly controlled computing
  1939. that financial institutions are capable of making. For start-ups innovating
  1940. in the computing space, such as tech start-ups, they may not be able
  1941. to consider the possibility. Highly controlled computer systems can
  1942. have very restrictive designs, and when these restrictions hinder the
  1943. innovation needed to implement the company&apos;s product, it will have
  1944. no choice but to pursue some other form of computing. After all, the company
  1945. rises or falls on the success of its innovation. That is not to say that
  1946. controlled enterprise computing is unimportant for such companies: quite
  1947. the contrary. The success of a start-up is highly dependent on a viable
  1948. ecosystem that provides known pathways to innovation while still moving
  1949. towards operating in a suitably controlled, production-ready way that
  1950. is necessary for any successful business. But for a technology start-up,
  1951. enterprise computing can never come at the expense of technological
  1952. innovation. The basic existence of the start-up company depends on its
  1953. ability to innovate: without innovation, there can be no company. In general,
  1954. this truth will hold in some form for any technology company, even well beyond
  1955. the start-up stage.
  1956.  
  1957. &lt;p&gt;The tension between innovation and control comes to the fore in
  1958. a different way at research-intensive universities, which are large
  1959. organizations with complex missions that need enterprise computing to
  1960. carry out their task of educating students on a broad scale, but are also
  1961. organizations committed to research, an activity that is, by its very
  1962. nature, an exploration into things not yet fully understood.  This conflict
  1963. is particularly acute in units within such universities that do research
  1964. into computing itself, such as computer science and computer engineering
  1965. departments, because in such places, the computer must serve both as the
  1966. locus of research and experimentation in addition to being a tool for
  1967. implementing institutional and departmental processes and the exercise of
  1968. legitimate control.
  1969.  
  1970. &lt;p&gt;I&apos;ve had the privilege of working in such a department, Computer
  1971. Science, at such a university (the University of Toronto) for more than
  1972. three decades now, most of that time in a computing leadership role,
  1973. and I know this tension all too well. It is sometimes exhausting, but
  1974. at the same time, it can also be a source of creative energy: yes, it is a
  1975. barrier, like a mountain athwart your path, but also, as a mountain to a
  1976. mountain-climber, a challenge to be overcome with determination, planning,
  1977. insight, and endurance. This challenge can be successfully overcome at
  1978. a good university, because in addition to a typical large organization&apos;s
  1979. commitment to basic values such as accountability, equity, reliability and
  1980. security, the university is equally committed to fundamental academic values
  1981. such as creativity, innovation and excellence. I look for ways to achieve
  1982. both. Over the years, I have had some successes. My department has produced
  1983. some groundbreaking research using academic computing that my technical staff
  1984. have been able to provide, and the department has been able to operate (and
  1985. successfully interoperate) in good cooperation with enterprise computing
  1986. at the divisional level, and with the central university as well.
  1987.  
  1988. &lt;p&gt;Yet I believe even more is possible. I have lived the tension in
  1989. both directions: to our researchers I at times have had to play the
  1990. regulator, having to impose constraints on computing to try to ensure
  1991. acceptable reliability, accountability and security. To our central
  1992. university computing organizations, I at times have had to advocate for
  1993. looser controls to create more room to innovate, sometimes in opposition to
  1994. proposals intended to increase reliability, security and accountability.
  1995. When things went badly, it was because one side or the other decided that
  1996. the other&apos;s concern is not their problem, and tried to force or sidestep the
  1997. issue. But when things went well, and most often it has, it is because
  1998. both sides genuinely recognized that at a research-intensive institution,
  1999. everyone needs to work within the tension between the need to innovate
  2000. and the need to regulate. As a body needs both a skeleton and flesh,
  2001. so too does a research university need both regulation and innovation:
  2002. without one, it collapses into a puddle of jelly; without the other,
  2003. into a heap of dry bones.
  2004.  
  2005. &lt;p&gt;With both being needed, one challenge to overcome is the fact that those
  2006. responsible for enterprise computing cannot be the same people responsible
  2007. for innovative research computing, and that is necessarily so. The
  2008. skill-sets vary, the domains vary, the user-base is quite different,
  2009. and the scale varies. If the university were to entrust both computing
  2010. innovation for computer science or computer engineering to the same
  2011. groups that provide enterprise computing for an entire large university,
  2012. one of two things would happen. Either the control necessary for a large
  2013. enterprise would be diminished in order to make room for innovation,
  2014. or, more likely, innovation would be stifled because of the need to
  2015. create sufficiently controlled enterprise computing at a suitable scale
  2016. for the entire university. Thus, necessarily, those who support unit
  2017. research computing, where the innovation takes place, will be different
  2018. people from those who support enterprise computing. But that can be a
  2019. strength, not a weakness. Rather than see each other as rivals, the two
  2020. groups can partner, embracing the tension by recognizing each others&apos;
  2021. expertise and each recognizing the others&apos; importance for the University
  2022. as a whole. Partnership brings many potential benefits: if innovation
  2023. becomes needed in new areas, for example, when the rise of data science
  2024. increasingly drives computing innovation outside of the traditional
  2025. computer science and computer engineering domains, the partnership can be
  2026. there to support it. Similarly, as the computing landscape shifts, and
  2027. new controls and new regulation becomes needed to address, for example,
  2028. emergent threats in information security, the partnership can be there to
  2029. support it. There is no organization potentially better suited for such a
  2030. partnership than a large research university, which, unlike a financial
  2031. institution, is profoundly committed to research and innovation through
  2032. its academic mission, but also, unlike a start-up, is a large and complex
  2033. institution with deep and longstanding responsibilities to its students,
  2034. faculty and community, obligated to carry out the enterprise computing
  2035. mission of accountability, reliability and security.
  2036.  
  2037. &lt;p&gt;So what might a partnership look like? It can take a number of different
  2038. forms, but in my view, whatever form it takes, it should have three key characteristics:
  2039. &lt;ul&gt;
  2040. &lt;li&gt;Locality&lt;/li&gt;
  2041. &lt;li&gt;Respectful Listening&lt;/li&gt;
  2042. &lt;li&gt;Practical Collaboration&lt;/li&gt;
  2043. &lt;/ul&gt;
  2044.  
  2045. &lt;p&gt;&lt;em&gt;Locality&lt;/em&gt; means that the computing people responsible for
  2046. research computing must stay close to the researchers who are
  2047. innovating. This is necessary for strictly practical reasons: all
  2048. the good will in the world is not enough to make up for a lack of knowledge
  2049. of what is needed most by researchers at a particular time. For example,
  2050. &lt;em&gt;deep learning&lt;/em&gt; is the dominant approach in Artificial Intelligence
  2051. today because a few years ago, our technical staff who supported research
  2052. computing worked very closely with researchers who were pursing deep
  2053. learning research, customizing the computing as necessary to meet the
  2054. research needs. This not only meant that we turned graphics cards into
  2055. computation engines at a time when this was not at all common and not
  2056. yet up to enterprise standards of reliability, it even means that at one
  2057. point we set up a research computer in a researcher&apos;s bedroom so that
  2058. he could personally watch over a key computing job running day and night
  2059. for the better part of a week. While this sort of customizability is not
  2060. always needed, and sometimes is not even possible (one could never run
  2061. a large computer centre this way), being able to do it if necessary is
  2062. a key research asset. A university will never be able to fully support
  2063. research computing solely from a central vantage-point. A commitment to
  2064. ensuring local presence and support of research computing operating at
  2065. the researcher level is necessary.
  2066.  
  2067. &lt;p&gt;&lt;em&gt;Respectful Listening&lt;/em&gt; means that the computing people responsible
  2068. for research computing at the unit level where research actually happens,
  2069. and the people responsible for enterprise computing divisionally and
  2070. centrally must communicate frequently, with an up-front commitment to hear
  2071. what the other is saying and take it into account. When problems arise,
  2072. respectful listening means that those problems will not be &quot;solved&quot; by
  2073. simply overruling or ignoring the other, to pursue a simplistic solution
  2074. that suits only one side. It also means a profound commitment to stepping
  2075. away from traditional organizational authority structures: just because the
  2076. innovative computing is situated in a department and the enterprise computing
  2077. is lead from the centre should not mean the centre should force its view on
  2078. the department, just because it can. Similarly, just because unit research
  2079. computing is driven by research faculty who enjoy substantial autonomy
  2080. and academic freedom, their research computing group at the unit level
  2081. should not simply ignore or sidestep what the enterprise is saying, just
  2082. because it can. Rather, both sides need to respect the other, listening to,
  2083. not disregarding, the other.
  2084.  
  2085. &lt;p&gt;&lt;em&gt;Practical Collaboration&lt;/em&gt; means that enterprise computing and
  2086. unit research computing need to work together in a collaborative way that
  2087. respects and reflects the timelines and resource constraints of each
  2088. side. Centrally offered computing facilities should support and empower
  2089. research where they can, but in a practical way: it may not be possible to
  2090. make a central facility so flexible and customizable that all research
  2091. can be pursued. It is acceptable to capture some research needs without
  2092. feeling an obligation to support the entire &quot;long tail&quot; of increasingly
  2093. customized research projects. Unit research computing will need to recognize
  2094. that the need to scale a centralized computing service may constrain the
  2095. amount of customizability that may be possible. Similarly, unit research
  2096. computing should use, rather than duplicate, central services where it makes
  2097. sense, and run its own services where that makes sense. Both central and
  2098. unit research computing should recognize that there is a legitimate middle
  2099. ground where some duplication of services is going to occur: sometimes
  2100. the effort required to integrate a large scalable central service into
  2101. a smaller customizable research service is too great, and sometimes the
  2102. research advantages of having a locally-run standardized service on which
  2103. experiments can more easily be built, can more than outweigh any sort of
  2104. economies of scale that getting rid of the unit service in favour of a
  2105. central service could theoretically provide. Hence the collaboration must be
  2106. practical: rather than slavishly pursue principles, it must be realistic,
  2107. grounded, balanced, sensible. It should recognize that one size does not
  2108. always fit all, and responsibly and collaboratively allocate resources in
  2109. order to preserve the good of the research mission.
  2110.  
  2111. &lt;p&gt;It is that research mission, the ability to innovate, that can make
  2112. computing so transformative at a research university. Yet while innovative
  2113. computing can indeed produce transformative change, it cannot be any change,
  2114. and not at any cost. Computing is a change agent, yes, but it is also a
  2115. critical component in the maintenance of an organization&apos;s commitment to
  2116. reliability, accountability, equity, and good operation. Success is found
  2117. in the maintenance of a suitable balance between the need to innovate and
  2118. the need to control. When an organization critically depends on both
  2119. factors, as a research university invariably does, I believe collaborative
  2120. partnerships between respective computing groups is the best way to maintain
  2121. the balance necessary for success.
  2122. </description>
  2123.  </item>
  2124.  <item>
  2125.    <title>Why it is a good idea to get the Covid19 vaccine?</title>
  2126.    <pubDate>Sun, 12 Sep 2021 18:05:00 -0400</pubDate>
  2127.    <link>http://www.cs.toronto.edu/%252525252525252525257Ejdd/blog.cgi/2021/09/12#whyvaccine</link>
  2128.    <category>/misc</category>
  2129.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525257Ejdd/blog.cgi/misc/whyvaccine</guid>
  2130.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  2131. &lt;IMG src=&quot;/~jdd/blog/misc/virus-1812092_1280.jpg&quot; alt=&quot;Visual representation of Covid19 viruses under electron microscope&quot;&gt;
  2132. &lt;figcaption&gt;&lt;small&gt;
  2133. Image by &lt;a href=&quot;https://pixabay.com/users/qimono-1962238/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm
  2134. _campaign=image&amp;amp;utm_content=1812092&quot;&gt;Arek Socha&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;
  2135. amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=1812092&quot;&gt;Pixabay&lt;/a&gt;
  2136. &lt;/small&gt;&lt;/figcaption&gt;
  2137. &lt;/figure&gt;
  2138.  
  2139. Like many people, I&apos;ve been following coverage of the COVID19 pandemic on the
  2140. news. One thing that seems to be coming up more and more is vaccine refusal:
  2141. some people are choosing not to get vaccinated for COVID19. Most people
  2142. around me know very well the benefits of vaccination. For them,
  2143. this vaccine refusal is idiotic: why would you not be vaccinated against
  2144. a disease that spreads very easily, and could kill you (and/or others)
  2145. or leave you (and/or others) with permanent health problems? They are
  2146. exasperated and annoyed at those who decline vaccination.
  2147.  
  2148. &lt;p&gt;While I understand becoming short of patience with vaccine refusal,
  2149. I don&apos;t think that most people who refuse COVID19 vaccination are
  2150. idiots. Vaccination and viruses can be complicated to understand. There are
  2151. a lot of misinformed posts and videos on the Internet. If you don&apos;t
  2152. know enough about how viruses and vaccines work, both in general and for
  2153. COVID19, how would you know what to believe? When my father died
  2154. of COVID19 last summer, one of the ways I dealt with the loss was
  2155. through understanding better how COVID19 works and what can be
  2156. done to fight it. My hope here is that by explaining the benefits
  2157. of vaccination in simple terms, I can maybe help others avoid
  2158. COVID19. I hope you will find it helpful. If not, there are &lt;A
  2159. HREF=&quot;https://www.canada.ca/covid-vaccines/q&amp;a&quot;&quot;&gt;other sites that address
  2160. this same question: maybe you will like those better?&lt;/A&gt;
  2161.  
  2162. &lt;p&gt; It all comes out of how viruses work. Viruses are not alive themselves,
  2163. but they use the bodies of living creatures (like us) to spread. They
  2164. find their way into the cells of our body, then take over those cells to
  2165. produce more copies of themselves. This is the only way viruses spread:
  2166. they can&apos;t reproduce on their own. For COVID19, you may have heard of the
  2167. &quot;spike protein&quot;. This is the spiky part on the outside of the COVID19 virus
  2168. that makes it look like a spiky ball. It&apos;s why it&apos;s called a &quot;coronavirus&quot;,
  2169. it looks a little like the spikes on a crown: &quot;corona&quot; is crown in
  2170. Latin. This protein helps the viruses get inside the body&apos;s cells. Then,
  2171. when inside, the viruses take over the cell to make and release more
  2172. viruses. Those viruses invade other cells, and those start making more
  2173. viruses too. Things can get out of hand very quickly, a bit like a
  2174. forest fire spreading in a dry forest.
  2175.  
  2176. &lt;p&gt; Happily, our body has a defence system against viruses (the &quot;immune
  2177. system&quot;). When those defences recognize an invading virus, it takes as
  2178. many viruses as possible out of action, keeping them from invading more
  2179. cells. If the defences can keep up, the viruses won&apos;t spread spread very
  2180. far, and our body will have fought off the infection. If the defences
  2181. can&apos;t keep up, the infection spreads.
  2182.  
  2183. &lt;p&gt; But our body&apos;s immune system needs to know, first, that something it sees
  2184. is a virus, before it can act. Immune systems learn from exposure and time.
  2185. If the body is exposed to enough viruses over time, the immune system can
  2186. learn how to recognize the virus, and start fighting back. When someone
  2187. gets sick from a viral infection like COVID19, they get sick because the
  2188. virus is spreading faster than the immune system can fight it off. Because
  2189. the immune system needs time to learn how to recognize the virus, while
  2190. it is learning, the virus is spreading, faster and faster. Sadly, this
  2191. can cause significant damage, depending on how far ahead the virus gets.
  2192. This is what happened to my father last summer when he caught COVID19.
  2193. At first, It spread much faster than his body could fight it, because his
  2194. immune system had to first learn how. As COVID19 spread, it caused damage
  2195. to his organ systems, including his heart. When his body&apos;s defences finally
  2196. learned how to fight off COVID19, the damage it had already done to his
  2197. heart was too great for him to stay alive. Sadly, he passed away shortly after.
  2198.  
  2199. &lt;p&gt; If the body survives, its immune system can remember viruses that it
  2200. has learned to recognize. When it is exposed later to the same
  2201. virus, it recognizes it right away, and fights it off quickly before it
  2202. can spread. This is why if you have successfully recovered from a viral
  2203. disease, you are less likely to get it later. This is the basis of
  2204. vaccination. Vaccination trains the body&apos;s immune system to recognize a virus
  2205. quickly, so that it will be able to muster a strong defence against it
  2206. right away, without giving the virus much chance to spread.
  2207.  
  2208. &lt;P&gt; The way COVID19 vaccinations work is that they train the body&apos;s
  2209. immune system to recognize the spike protein on the outside of a COVID19
  2210. virus. It doesn&apos;t inject the spike protein itself, but rather it injects
  2211. something that instructs the body&apos;s cells to temporarily produce a bit
  2212. of spike protein for training. Your body&apos;s defences learns from this to
  2213. recognize anything with the spike protein (such as a COVID19 virus) as an
  2214. invader. If later it is exposed to COVID19 virus, the body&apos;s defences will
  2215. be primed and ready to get rid of it before it can spread very far.
  2216.  
  2217. &lt;p&gt; Unfortunately, the body&apos;s defences against viruses aren&apos;t perfect. In
  2218. the case of COVID19, a single exposure to the spike protein does train the
  2219. body to recognize it, but not always quickly and thoroughly enough. Like
  2220. us, when we&apos;re learning a new skill, our immune systems learn better with
  2221. multiple lessons. That is why most COVID19 vaccinations require two shots:
  2222. the immune system learns better with two lessons than one, and in some cases
  2223. three (a booster) rather than two. This is also why people who have had
  2224. COVID19 should still get vaccinated: a successful recovery from a COVID19
  2225. infection does provide some protection, but additional lessons for the
  2226. body&apos;s defences will still help if exposed to the virus again. This is
  2227. also the reason why vaccinations are not perfect. They train the
  2228. body&apos;s immune system to recognize and eliminate the virus, but if the body
  2229. is exposed to too much virus too quickly, the viruses can still spread
  2230. faster than the immune system can eliminate it. This is why a few people
  2231. who are fully vaccinated do get sick from COVID19, though not usually as
  2232. seriously as people who were not vaccinated. This doesn&apos;t mean that the
  2233. vaccine &quot;doesn&apos;t work&quot;, it just means that even trained immune systems
  2234. can sometimes be overwhelmed by a virus.
  2235.  
  2236. &lt;p&gt; Because vaccinations train the immune system to
  2237. recognize and fight off a virus, after a vaccination, &lt;A
  2238. HREF=&quot;https://www.mcgill.ca/oss/article/covid-19-health/i-felt-sick-after-getting-vaccine-why&quot;&gt;we
  2239. sometimes feel a bit sick&lt;/A&gt;: some of the symptoms we experience when we
  2240. are sick are caused by the body&apos;s defences: e.g. fever,
  2241. aches, fatigue,and feeling unwell. In the case of a vaccination, though,
  2242. this is not long-term, because a vaccination, unlike a virus, does not
  2243. reproduce and spread, and so its effects will wear off quickly.
  2244.  
  2245. &lt;p&gt; Vaccinations can sometimes cause side effects that are more serious. This
  2246. is why they are tested carefully before approval. In the case of the major
  2247. COVID19 vaccines, there are some very rare side effects that are serious: &lt;A
  2248. HREF=&quot;https://covid19-sciencetable.ca/sciencebrief/risk-of-vaccine-induced-thrombotic-thrombocytopenia-vitt-following-the-astrazeneca-covishield-adenovirus-vector-covid-19-vaccines/&quot;&gt;certain
  2249. COVID19 vaccines cause very rare but quite serious blood clots&lt;/A&gt;, and &lt;A
  2250. HREF=&quot;https://www.cdc.gov/coronavirus/2019-ncov/vaccines/safety/myocarditis.html&quot;&gt;certain
  2251. others cause very rare heart inflammation&lt;/A&gt;. These side-effects don&apos;t happen very
  2252. often in people who receive the vaccine: they are much less likely than the
  2253. &lt;A HREF=&quot;https://www.nationalgeographic.com/science/article/flash-facts-about-lightning&quot;&gt;probability of the average person being hit by lightning in their lifetime&lt;/A&gt;.
  2254.  
  2255. &lt;p&gt; The fact is, the vaccine is much less dangerous than
  2256. the disease. A COVID19 infection can cause very serious
  2257. health effects, and many of those effects are not rare.
  2258. While most people who catch COVID19 recover at home, more than &lt;A
  2259. HREF=&quot;https://www.sciencedirect.com/science/article/pii/S0002934321000991&quot;&gt;one
  2260. in twenty&lt;/A&gt; require hospitalization to stay alive. Of those, on the order of
  2261. &lt;A HREF=&quot;https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7920817/&quot;&gt;one
  2262. in ten die&lt;/A&gt;.  Moreover, many who recover from COVID19 &lt;A
  2263. HREF=&quot;https://www.mayoclinic.org/diseases-conditions/coronavirus/in-depth/coronavirus-long-term-effects/art-20490351&quot;&gt;suffer
  2264. long-term health effects ranging from difficulty breathing, to fatigue, pain,
  2265. and memory, concentration and sleep problems. Organ damage to the heart,
  2266. lungs and brain is also possible. &lt;/A&gt;COVID19 is spreading around the world
  2267. and most people will eventually be exposed to it. It is better to get the
  2268. vaccine first, so that you are less likely to be harmed by the disease later.
  2269.  
  2270. &lt;p&gt; There are claims on the Internet that COVID19 vaccines are much
  2271. more dangerous than what I&apos;ve written here. Many of these claims are
  2272. misunderstandings. Millions of people have received COVID19 vaccines.
  2273. A few who have had health problems after receiving the vaccine have reported
  2274. their health problems as a possible &quot;side effect&quot; of the virus. In the
  2275. US, there is a &lt;A HREF=&quot;https://vaers.hhs.gov&quot;&gt;vaccine reporting system
  2276. called VAERS&lt;/A&gt; where people can report bad health events that happened to
  2277. them after receiving a vaccine: this lets scientists investigate whether
  2278. the vaccine might have caused the problem.  If the vaccine is causing
  2279. a particular health problem, that problem would happen more often to
  2280. people who receive the vaccine than to those who do not. But for most
  2281. of the health problems reported to VAERS, they are not happening more
  2282. often to vaccinated people, they happen at roughly the same rate as they
  2283. happen to anyone, and so the vaccine cannot be responsible. It appears
  2284. that COVID19 vaccines cause very few serious health problems, and those
  2285. are very rare. The evidence for this is that millions of people around the
  2286. world have received COVID19 vaccines and almost nobody has gotten seriously
  2287. sick from them. The COVID19 disease itself is much more dangerous, which is
  2288. why hospitals are full of people suffering from the disease, not the vaccine.
  2289.  
  2290. &lt;p&gt; Even so, wouldn&apos;t it be better to avoid both the vaccine and the
  2291. disease? Yes, it would be, if you could be assured
  2292. of never being exposed to COVID19. But that is not so easy. &lt;A
  2293. HREF=&quot;http://www.bccdc.ca/health-info/diseases-conditions/covid-19/about-covid-19/how-it-spreads&quot;&gt;COVID19
  2294. spreads very easily: it spreads through tiny moisture droplets in exhaled
  2295. breath&lt;/A&gt; that float in the air like smoke from a cigarette, so if you
  2296. are indoors with someone who is exhaling COVID19 virus, and there is
  2297. poor air circulation, you will inhale some. The longer
  2298. you are there, the more COVID19 virus you will inhale. &lt;A
  2299. HREF=&quot;https://www.nature.com/articles/d41586-021-02259-2&quot;&gt;Not everyone
  2300. who gets COVID19 feels very sick right away: some feel fine, at least
  2301. for a while&lt;/A&gt;, and many who feel sick don&apos;t feel so sick that they stay home.
  2302. They will spread the virus whereever they go, simply by exhaling. You may be in a room with
  2303. an infected person who has no idea that they are spreading COVID19. This is
  2304. why masks are so helpful, because the mask over the nose and mouth of an
  2305. infected person reduces the amount of COVID19 viruses they breathe out,
  2306. and the mask over the nose and mouth of other people in the room reduces
  2307. the amount of COVID19 virus they might breathe in. It&apos;s also a reason why
  2308. indoor fresh air circulation is so important, and why COVID19 is so much more
  2309. of a danger indoors than outdoors. COVID19 is very contagious, especially
  2310. the new &quot;delta&quot; variant which is the dominant variant circulating today: &lt;A
  2311. HREF=&quot;https://www.npr.org/sections/goatsandsoda/2021/08/11/1026190062/covid-delta-variant-transmission-cdc-chickenpox&quot;&gt;on
  2312. average, a sick person will spread it to six or more others. It&apos;s only a
  2313. little less transmissible than chickenpox, but a lot more transmissible than
  2314. flu.&lt;/A&gt; It&apos;s quite possible that we will all be exposed to it eventually.
  2315.  
  2316. &lt;p&gt;An even more important reason to be vaccinated is to reduce the spread of
  2317. COVID19 to others. Remember that the only way for a virus to reproduce
  2318. is in the body of an infected person. If most people make their bodies
  2319. inhospitable to the virus by getting vaccinated, then the virus will find
  2320. very few opportunities to spread. It&apos;s like fire trying to spread in a
  2321. very wet forest: only the dry sticks will burn, and the fewer dry sticks
  2322. there are, the less likely the fire will find more sticks to spread to, and
  2323. the more likely it will burn out. So by getting vaccinated, we protect not
  2324. only ourselves, but everyone around us, especially those who, for medical
  2325. reasons, can&apos;t be vaccinated, or who have immune systems that don&apos;t work
  2326. well. If not enough of us get vaccinated, the number of COVID19 cases will
  2327. overwhelm the hospitals. Most of those who need hospital care for their
  2328. COVID19 infections will die instead. Also, many people who need hospital
  2329. care for other serious illnesses won&apos;t be able to get the care they need,
  2330. and they will die too.
  2331.  
  2332. &lt;p&gt;So please be brave: if you can, get vaccinated. Yes, the effects of
  2333. the vaccine may be unpleasant for a few days as the body learns how to
  2334. fight the virus. But the vaccine will not harm you like the disease will,
  2335. and it will train the body&apos;s immune system to fight it. My father got COVID19 too early, last
  2336. summer, before COVID19 vaccines were available. If they had been available
  2337. then, he might still be alive today. They&apos;re available now. Please get
  2338. vaccinated if you can. If enough people around the world get vaccinated
  2339. against COVID19, we may eventually be able to eliminate this disease
  2340. altogether, and that would be a thing worth doing.
  2341. </description>
  2342.  </item>
  2343.  <item>
  2344.    <title>The Covid19 Blues</title>
  2345.    <pubDate>Sun, 06 Jun 2021 13:39:00 -0400</pubDate>
  2346.    <link>http://www.cs.toronto.edu/%25252525252525252525257Ejdd/blog.cgi/2021/06/06#covidblues</link>
  2347.    <category>/misc</category>
  2348.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525257Ejdd/blog.cgi/misc/covidblues</guid>
  2349.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  2350. &lt;IMG src=&quot;/~jdd/blog/misc/guitar-3291890_1280.jpg&quot; alt=&quot;Man playing a guitar&quot;&gt;
  2351. &lt;figcaption&gt;&lt;small&gt;
  2352. Image by &lt;a href=&quot;https://pixabay.com/users/lucasvieirabr-8401756/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3291890&quot;&gt;lucasvieirabr&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3291890&quot;&gt;Pixabay&lt;/a&gt;
  2353. &lt;/small&gt;&lt;/figcaption&gt;
  2354. &lt;/figure&gt;
  2355.  
  2356. &lt;p&gt;The arts find inspiration in times of trouble, none more so than the sort of
  2357. music known as the blues. Blues are creative and emotional,
  2358. sometimes raw, but never fake. Blues are not about superstars
  2359. and megahits, blues are about the endurance and hope of ordinary
  2360. people. As Covid19 drags on, endurance and hope are needed more than
  2361. ever. Here are pointers to a few Covid19-inspired blues
  2362. tracks that I appreciate.
  2363.  
  2364. &lt;ul&gt;
  2365. &lt;li&gt; &lt;A HREF=&quot;https://www.youtube.com/watch?v=MfNpa372t-Y&quot;&gt;Keb&apos; Mo&apos;&lt;/A&gt;
  2366. &lt;li&gt; &lt;A HREF=&quot;https://www.youtube.com/watch?v=4bkPOgHRXPw&quot;&gt;Rob Bentley&lt;/A&gt;
  2367. &lt;li&gt; &lt;A HREF=&quot;https://www.youtube.com/watch?v=1H31SiJC-8Q&quot;&gt; Mads Jacobsen&lt;/A&gt;
  2368. &lt;li&gt; &lt;A HREF=&quot;https://www.youtube.com/watch?v=hwHkYiPIZsk&quot;&gt; (Mads has an &quot;Astronaut Remix&quot; version too)&lt;/A&gt;
  2369. &lt;li&gt; &lt;A HREF=&quot;https://twitter.com/KarlKohlhase/status/1388477942192279556&quot;&gt; Karl Kohlhase&lt;/A&gt;
  2370. &lt;/ul&gt;
  2371.  
  2372. Enjoy!
  2373. </description>
  2374.  </item>
  2375.  <item>
  2376.    <title>What&apos;s Wrong With Passwords on the Internet Anyway?</title>
  2377.    <pubDate>Thu, 31 Dec 2020 22:57:00 -0500</pubDate>
  2378.    <link>http://www.cs.toronto.edu/%2525252525252525252525257Ejdd/blog.cgi/2020/12/31#passwords</link>
  2379.    <category>/it</category>
  2380.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525257Ejdd/blog.cgi/it/passwords</guid>
  2381.    <description>
  2382. &lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  2383. &lt;IMG src=&quot;/~jdd/blog/it/registration-3938434_1280.jpg&quot; alt=&quot;Completed Login Prompt&quot;&gt;
  2384. &lt;figcaption&gt;&lt;small&gt;
  2385. Image by &lt;a href=&quot;https://pixabay.com/users/geralt-9301/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3938434&quot;&gt;Gerd Altmann&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=3938434&quot;&gt;Pixabay&lt;/a&gt;
  2386. &lt;/small&gt;&lt;/figcaption&gt;
  2387. &lt;/figure&gt;
  2388.  
  2389. More than fifteen years ago, &lt;A HREF=&quot;https://www.cnet.com/news/gates-predicts-death-of-the-password/&quot;&gt;Bill Gates predicted that use of
  2390. traditional passwords would dwindle.&lt;/a&gt; This has happened to a certain extent, but a login and password is still the most
  2391. commonly used credential for computing authentication. It is
  2392. increasingly problematic. According to &lt;A
  2393. HREF=&quot;https://enterprise.verizon.com/resources/reports/2020-data-breach-investigations-report.pdf&quot;&gt;
  2394. Verizon&apos;s 2020 Data Breach Investigations report, 37% of all
  2395. breaches involved the stealing of credentials or the use of stolen
  2396. credentials. (p.7)&lt;/A&gt; What is the root cause of the problem?
  2397.  
  2398. &lt;p&gt; Put in simple terms, a login and password is what a system relies on
  2399. to know who is who. Your password is secret: only you know what it is,
  2400. and the system has some way of checking that it is correct. If someone
  2401. connects to the system with your login and password, the system checks
  2402. that the password is the right one for your login. If it is, the system
  2403. concludes that you are the person trying to connect, and lets you in.
  2404. If you are the only one who knows the password, this approach works,
  2405. since you are the only person who can provide the correct password. But
  2406. if criminals know your password too, and use it, the system will think the
  2407. criminals are you, and will give them access to your account and all your
  2408. data. The only way to fix this is to change your password to something
  2409. new that only you know, but by then the damage may well be done.
  2410.  
  2411. &lt;p&gt; Unfortunately, criminals have a pretty effective technique for finding
  2412. out your login and password: they trick you into telling it to them. &quot;Wait
  2413. a minute!&quot;, you might say, &quot;I won&apos;t ever tell a criminal my password. I don&apos;t
  2414. even tell my family my password!&quot; But you tell the system your password
  2415. every time you log in. So if criminals set up a fake system that looks
  2416. like the real one, and trick you into trying it, when you tell their fake
  2417. system your password, the criminals will learn what it is.
  2418.  
  2419. &lt;p&gt; This was not a common problem in the past, because it was difficult
  2420. for criminals to successfully set up fake systems that look convincing.
  2421. But on the Internet today, it is easy to set up a web site that looks like
  2422. another site. The only thing that&apos;s hard to fake is the first part
  2423. of the link, the hostname section that comes immediately after the double
  2424. slash (//) and before the first single slash (/), because that part of the
  2425. link is used to direct the request to the right system on the Internet. But
  2426. given that the Internet is available in hundreds of countries, each with
  2427. its own set of internet service providers, it is often not too difficult
  2428. for criminals to find somewhere on the Internet where they can register a
  2429. hostname that is similar-looking to the real thing.
  2430.  
  2431. &lt;p&gt; Worse, the rise of messages containing embedded links make it very easy
  2432. for criminals to send a fake message (e.g. an email or text) with a link that
  2433. seems legitimate but really directs you to a fake site. This is
  2434. called &quot;phishing&quot;. Because of the way the web&apos;s markup language
  2435. ( HTML) works, it is easy to set up a link that seems to point
  2436. to one site, but actually points to another. For example, &lt;A
  2437. HREF=&quot;https://www.amazon.com&quot;&gt;https://www.walmart.com&lt;/A&gt; is a link that
  2438. seems to point to Walmart but really points to Amazon. Most web browsers
  2439. will let you &quot;hover&quot; over a link to see where it really goes.  But do
  2440. people check every link carefully each time they use it?
  2441.  
  2442. &lt;p&gt;The problem is made worse by the proliferation of legitimate messages with
  2443. embedded links to all sorts of cloud services. I recently saw a message
  2444. from a large organization to its staff, about their pensions. The message
  2445. contained links to an external site whose name had no resemblance to
  2446. the organization&apos;s name. The message invited the staff to click on those
  2447. links to see information about their pensions. The message was legitimate:
  2448. the organization had contracted with an external cloud provider to provide
  2449. an online pension calculator for staff. But the message said nothing
  2450. about the cloud provider: it merely contained a link to the calculator.
  2451. If criminals had sent a similar message containing a malicious link to a
  2452. fake system somewhere on the Internet, one that prompted staff to enter
  2453. their login and password, no doubt many staff would have thought it
  2454. legitimate. How could staff be expected to be able to tell the difference?
  2455.  
  2456. &lt;p&gt;A good way to combat the password capturing problem is to require
  2457. more than just a password to use a system. This is called &quot;two-factor&quot; or
  2458. &quot;multi-factor&quot; authentication. Your password is one factor, and something
  2459. else is a second factor, and you must provide both factors to prove to the
  2460. system that it is you. This helps because the criminals must have both your
  2461. password and your second factor in order to access your account and data.
  2462. To ease the authentication burden for users, systems can ask for two
  2463. factors only sometimes, such as when logging in for the first time in a
  2464. while, or logging in from a new machine or a new location.
  2465.  
  2466. &lt;p&gt; Ideally the second factor should be something that is hard for
  2467. criminals to capture and use. One problem with a password is that it is a
  2468. secret that can be used from anywhere on the Internet. With &lt;A
  2469. HREF=&quot;https://www.statista.com/statistics/617136/digital-population-worldwide/&quot;&gt;
  2470. almost 60% of the world&apos;s population on the Internet&lt;/A&gt;, which now
  2471. reaches every country in the world, the Internet can hardly be considered a
  2472. &quot;safe place&quot;. A second password, as easily used from anywhere on the
  2473. Internet as the first, would not be much of an improvement. Worse would be
  2474. the answers to some personal question about yourself, such as your mother&apos;s
  2475. maiden name or the name of your first school: not only is such information
  2476. just as easily used as a password, it is information that people may be
  2477. able to find out in various ways. Answers to personal questions, while
  2478. sometimes used for authentication, typically do not make a good second factor.
  2479.  
  2480. &lt;p&gt; A better second factor is a message sent via a communication
  2481. channel that goes only to you: for example, an email to your email
  2482. address, or a text to your cell phone number. When you attempt to
  2483. log in, the system sends a unique one-time code to you through that
  2484. channel, and asks you to enter it. The assumption is that criminals
  2485. won&apos;t have access to your email or your cell number, so they won&apos;t know
  2486. and be able to enter the one-time code that the system sent to you. This
  2487. is usually a good assumption.  But criminals can try to get access
  2488. to your email or your phone number, and sometimes they succeed. For
  2489. example, in the case of a cell number, one thing they could try is to &lt;A
  2490. HREF=&quot;https://www.howtogeek.com/358352/criminals-can-steal-your-phone-number.-heres-how-to-stop-them/&quot;&gt;call
  2491. your cell phone provider, tell them they are you and that your phone has
  2492. been stolen, and request that your phone number be transferred to their
  2493. new phone.&lt;/A&gt;
  2494.  
  2495. &lt;p&gt; Another second factor, one even better, is a physical device in your
  2496. possession. This could be a hardware security token that you plug into
  2497. your computer or that displays a unique, frequently changing, code. Or it
  2498. could be an app on your cell phone that is tied to your unique device. A
  2499. physical device is an excellent second factor, because most criminals on
  2500. the Internet are physically distant. To successfully pretend to be you,
  2501. a criminal would need direct physical access to a device that would likely
  2502. be located in your purse or pocket.
  2503.  
  2504. &lt;p&gt; Relying on a device in purse or pocket as well as a password in your head
  2505. is an improvement in security, but it has its drawbacks. It makes that device
  2506. essential for you to use the system: if it is broken, lost or stolen,
  2507. you&apos;re locked out, even if you know the password. While locking out people
  2508. who don&apos;t have the device is exactly the point, that doesn&apos;t help when it is
  2509. keeping you from legitimately using the system. Moreover, if that device
  2510. is your smartphone, it changes your phone from a convenience to a
  2511. necessity. While a smartphone has become a necessity already to some,
  2512. it is a potentially consequential thing for it to become a requirement for
  2513. everyone. A hybrid approach is perhaps best: hardware security tokens those
  2514. who prefer it, a smartphone for those who for their own reasons carry one
  2515. around anyway, and for many, both: a smartphone for convenience, with a
  2516. hardware security token as backup, in case of smartphone loss or damage.
  2517.  
  2518. &lt;p&gt; Perhaps there is an even more secure option? What if your second
  2519. factor wasn&apos;t a device, but an actual physical part of your body,
  2520. such as a finger (for a fingerprint), eye (for a retinal scan), face,
  2521. or even heartbeat (as measured by e.g. a &lt;A HREF=&quot;https://www.nymi.com/&quot;&gt;
  2522. Nymi Band&lt;/A&gt;)? Would that be better still? After all, if it is hard
  2523. for a criminal to get access to someone&apos;s things without being noticed,
  2524. it is even harder to get access to someone&apos;s body. This is indeed
  2525. possible: a technique called &quot;biometrics, and it can be
  2526. an effective second factor. Unfortunately there are a couple of issues
  2527. with biometrics. For example, injuries or health issues can change your
  2528. body; a cut on your finger may affect your fingerprint, for instance.
  2529. Secondly, biometrics have a &quot;revocation&quot; problem. This comes from the fact
  2530. that a biometric is a unique measurement of your body part: a fingerprint,
  2531. retinal scan, facial image, or ECG. But measurements
  2532. are data, and biometric data, like any other data, can and &lt;A
  2533. HREF=&quot;https://www.forbes.com/sites/zakdoffman/2019/08/14/new-data-breach-has-exposed-millions-of-fingerprint-and-facial-recognition-records-report/&quot;&gt;
  2534. has been breached.&lt;/A&gt; If this happens, what will you do? Passwords can
  2535. be changed, hardware security tokens can be replaced, but how are you
  2536. going to change your fingerprint, your face, your eye, your heartbeat?
  2537. While biometrics do have a place in authentication, most commonly to unlock
  2538. a local device such as a smartphone or a laptop, the lack of revocability
  2539. make biometrics less suitable as a second factor for Internet-accessible
  2540. services.
  2541.  
  2542. &lt;p&gt; Regardless of what is chosen for a second factor, the inconvenience
  2543. of using more than one factor is something that has to be considered.
  2544. Passwords, especially ones that are easy to remember, are
  2545. quite convenient. Requiring more than this can make authentication more
  2546. difficult. If becomes too difficult, the difficulty becomes a disincentive
  2547. to use the system. For systems protecting highly sensitive data, some
  2548. difficulty may be warranted, given the risk. For lower-risk systems, things
  2549. are less clear. Yet for Internet-accessible systems, due to the prevalence
  2550. of phishing, something more secure than just passwords seems increasingly
  2551. necessary. I think Bill Gates is right: like it or not, the traditional
  2552. password will become increasingly rare on the Internet, for good reason.
  2553. </description>
  2554.  </item>
  2555.  <item>
  2556.    <title>Thoughts on Covid19</title>
  2557.    <pubDate>Mon, 23 Nov 2020 00:00:00 -0500</pubDate>
  2558.    <link>http://www.cs.toronto.edu/%252525252525252525252525257Ejdd/blog.cgi/2020/11/23#covid</link>
  2559.    <category>/misc</category>
  2560.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525257Ejdd/blog.cgi/misc/covid</guid>
  2561.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  2562. &lt;IMG src=&quot;/~jdd/blog/misc/virus-4937553_1280.jpg&quot; alt=&quot;Visual representation of Covid19 viruses under electron microscope&quot;&gt;
  2563. &lt;figcaption&gt;&lt;small&gt;
  2564. Image by &lt;a href=&quot;https://pixabay.com/users/PIRO4D-2707530/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=4937553&quot;&gt;PIRO4D&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=4937553&quot;&gt;Pixabay&lt;/a&gt;
  2565. &lt;/small&gt;&lt;/figcaption&gt;
  2566. &lt;/figure&gt;
  2567.  
  2568. I&apos;d recently reread a blog entry I&apos;d written more than a year ago now
  2569. on intentionality about blog posting. After writing it, I lived it: I wrote
  2570. several additional blog entries throughout the year. But then along came
  2571. the Covid19 pandemic, and it illustrated a problem with intentionality:
  2572. intentionality requires priority. When Covid19 hit Ontario in March, the pandemic
  2573. required substantial changes in how I live and work, and that drove a
  2574. reprioritization of my efforts, both in my job as Director responsible
  2575. for computing at the University of Toronto&apos;s Computer Science department,
  2576. and at home, as a parent of teenagers in high school.  In the face of
  2577. the challenges of Covid19, blogging seemed not sufficiently important,
  2578. and of course, it wasn&apos;t.  So I didn&apos;t write, I worked. I am grateful to
  2579. have work, in fact: I know of others who couldn&apos;t work because the sort
  2580. of work they did couldn&apos;t be done from home.  I consider myself fortunate
  2581. to work in the computing field, which has not been so badly affected. In
  2582. fact, in many ways, computing has been part of the solution (networking,
  2583. videoconferencing, cloud computing, medical informatics, etc.) and has been
  2584. boosted rather than impaired. In my job, I and my staff, and my department,
  2585. found ourselves not without work, but with too much. This is not necessarily
  2586. a bad situation to be in, but it doesn&apos;t lend itself to blogging.
  2587.  
  2588. &lt;p&gt; Another reason is that Covid19 didn&apos;t just affect me professionally, it
  2589. affected me personally: I lost a parent to Covid19 this summer.
  2590. While I am not in any way unique in having lost someone to this disease,
  2591. I was not really in a good state to blog, for quite some time.
  2592.  
  2593. &lt;p&gt; There is still another factor, though, one that also kept me from
  2594. blogging.  I am no epidemiologist. Still, as a thinking person, I seek to
  2595. understand what was going on, why, and what can be done about it.
  2596. Seeking to understand is, for me, theraputic: it helps me deal with stress,
  2597. anxiety, grief, and loss.
  2598.  
  2599. &lt;p&gt; First, I looked for good sources of information about the
  2600. pandemic itself. The &lt;A HREF=&quot;https://www.cdc.gov&quot;&gt; Centre
  2601. for Disease Control and Prevention&lt;/A&gt; in the US has plenty of
  2602. good material about it.  One thing I found particularly helpful was an &lt;A
  2603. HREF=&quot;https://www.cdc.gov/mmwr/volumes/69/wr/mm6919e6.htm&quot;&gt;analysis in
  2604. mid-May&lt;/A&gt; about a choir practice in Washington state with 61 attendees, one
  2605. that led to most becoming infected. It resulted in three hospitalizations and
  2606. two deaths. The CDC report is a very helpful example of rigorous statistical
  2607. data analysis set in a small, understandable real-world context. As an
  2608. illustration of what the Covid19 virus is like, I find it very helpful. For
  2609. instance, it suggested airborne spread before that became generally realized.
  2610.  
  2611. &lt;p&gt; Secondly, information about previous pandemics. Again,
  2612. the &lt;A HREF=&quot;https://www.cdc.gov&quot;&gt;Centre for
  2613. Disease Control and Prevention&lt;/A&gt; in the US has a &lt;A
  2614. HREF=&quot;https://www.cdc.gov/flu/pandemic-resources/basics/past-pandemics.html&quot;&gt;very
  2615. good past pandemics page&lt;/A&gt;, put together before the Covid19 pandemic
  2616. started, covering the horrifying 1918 influenza pandemic that killed fifty
  2617. million people around the world, and the later influenza epidemics of 1957,
  2618. 1968, and 2009. Each of these provide a general helpful picture: firstly,
  2619. that each pandemic has a timeframe that is typically greater than one year
  2620. but less than two, that transmission reduces in the summer but increases
  2621. in the fall/winter due to indoor crowding and decreased relative humidity,
  2622. and that mass vaccination can be an effective way to ward off a disaster
  2623. of the scale of the 1918 pandemic.
  2624.  
  2625. &lt;p&gt; One problem with this current pandemic is that, unlike the pandemics of
  2626. 1957, 68, and 2009, the virus is not influenza, but a coronavirus. There
  2627. are four coronaviruses that have been circulating widely for years &lt;A
  2628. HREF=&quot;https://www.cdc.gov/coronavirus/types.html&quot;&gt;(229E, NL63, OC43, and
  2629. HKU1&lt;/A&gt;), but they typically don&apos;t cause serious illness.  Two others
  2630. (SARS-CoV and MERS-CoV) emerged in the early 21st century, both quite
  2631. dangerous and certainly serious enough to warrant vaccination were they
  2632. to spread widely, but due to a great deal of diligence and effort, and not
  2633. a little good fortune, both of these were kept from spreading through
  2634. the world population. The current Covid19 pandemic,
  2635. caused by yet another coronavirus, SARS-CoV2, is the first coronavirus
  2636. both serious enough and widespread enough to warrant a vaccine. Unlike
  2637. for influenza, a coronavirus vaccine has never been produced before, so it has
  2638. taken longer than it would if this pandemic had been influenza. Only now,
  2639. as we approach the one year mark of the virus&apos; first emergence, are we
  2640. seeing some likely vaccine candidates. It will still take some time to
  2641. produce and distribute suitable vaccines.
  2642.  
  2643. &lt;p&gt; In the meantime, while efforts continue to design,
  2644. test, produce and distribute a suitable vaccine,
  2645. the challenge is to keep Covid19 from spreading far and
  2646. fast. While at first it was believed that &lt;A
  2647. HREF=&quot;https://www.canada.ca/en/public-health/services/diseases/2019-novel-coronavirus-infection/health-professionals/main-modes-transmission.html&quot;&gt;Covid19
  2648. spreads primarily through surface contact, there is increasing evidence
  2649. for areosol spread (fine droplets in the air)&lt;/A&gt;. So methods are needed
  2650. to hinder the passing of the virus from one person to another. There
  2651. are two main approaches: keeping people further apart, and putting physical
  2652. barriers (e.g. masks) and processes (e.g. handwashing) in place so that
  2653. the virus can&apos;t easily pass from one person to another.
  2654. &lt;p&gt;
  2655. The best way to hinder the transmission of Covid19 is to find out who may
  2656. be contagious (through testing and contact-tracing), and keep
  2657. them away from everyone else (quarantine) until they are no longer contagious.
  2658. One challenge is that it can sometimes be very hard to detect
  2659. when someone has Covid19 and is spreading the virus. There
  2660. is a wide variation in how Covid19 affects people who have it. For
  2661. many, it can take days for symptoms to emerge (presymptomatic), and &lt;A
  2662. HREF=&quot;https://www.healthline.com/health-news/20-percent-of-people-with-covid-19-are-asymptomatic-but-can-spread-the-disease&quot;&gt;for
  2663. some, Covid19 can be mostly or completely asymptomatic, yet asymptomatic and
  2664. presymptomatic Covid19 patients can spread the disease&lt;/A&gt;.  If those who
  2665. may have Covid19 can be identified (through testing and thorough contact
  2666. tracing), then those individuals alone can be quarantined until they are
  2667. no longer contagious. If they cannot be identified, then the only way
  2668. to hinder the spread of the disease is to assume that almost anyone might
  2669. have Covid19. This requires such things as requiring everyone to wear masks,
  2670. and, despite severe social and economic cost, lockdowns, which are a sort of
  2671. semi-quarantine for everyone. As I write this, Covid19 has been spreading
  2672. quite quickly in my city, Toronto, despite a mask mandate, and so &lt;A
  2673. HREF=&quot;https://www.toronto.ca/news/city-of-toronto-supports-provincial-lockdown-measures-to-reduce-covid-19-transmission/&quot;&gt;Toronto
  2674. is going into lockdown&lt;/a&gt;.  &lt;p&gt; How will it all end? In the struggle between
  2675. pessimism and hope, I choose hope. I hope that I will not lose any more
  2676. family members to this disease. I hope that effective vaccines will soon
  2677. be available in the necessary quantities. I hope that the measures taken
  2678. to hinder the spread will be effective. I think it is reasonable to expect
  2679. that we will see the widespread distribution of effective vaccines in 2021,
  2680. and this pandemic will be over sometime next year. Will everything be the
  2681. same? No, I think not. Some businesses (tourism and travel, for example)
  2682. will have a massive economic hole to climb out of, and some companies will
  2683. not survive, but people will travel again. Working from home, and technology
  2684. in support of it, will be more widely accepted. Cheek-to-jowl &quot;open-concept&quot;
  2685. offices, handshaking, and other close-quarters working practices will
  2686. be less readily accepted. There will be a greater consciousness of viral
  2687. hygiene, and a greater acceptance of masks. But life will go on. Covid19
  2688. will no longer command the attention it is getting now. Other things will
  2689. seem important again. And there will be many worthwhile things to blog about.
  2690. </description>
  2691.  </item>
  2692.  <item>
  2693.    <title>Some Clarity on Public Cloud Cybersecurity</title>
  2694.    <pubDate>Mon, 24 Feb 2020 10:19:00 -0500</pubDate>
  2695.    <link>http://www.cs.toronto.edu/%25252525252525252525252525257Ejdd/blog.cgi/2020/02/24#cloudsecurity</link>
  2696.    <category>/it</category>
  2697.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525257Ejdd/blog.cgi/it/cloudsecurity</guid>
  2698.    <description>&lt;figure style=&quot;float:left;width:40%;padding:1px&quot;&gt;
  2699. &lt;IMG src=&quot;/~jdd/blog/it/clear-114143_640.jpg&quot; alt=&quot;Break in clouds, revealing clear skies&quot;&gt;
  2700. &lt;figcaption&gt;&lt;small&gt;
  2701. Image by &lt;a href=&quot;https://pixabay.com/users/36882-36882/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=114143&quot;&gt;Sabrina Corana&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=114143&quot;&gt;Pixabay&lt;/a&gt;
  2702. &lt;/small&gt;&lt;/figcaption&gt;
  2703. &lt;/figure&gt;
  2704.  
  2705. I&apos;ve been thinking about public cloud cybersecurity for some years now,
  2706. as I&apos;ve watched adoption of the public cloud grow from a trickle to a
  2707. flood. Early on, most of the reasons I heard for public cloud adoption
  2708. made a great deal of sense to me: the need to rapidly scale up and
  2709. down the size of a service, the desire to leverage the expertise
  2710. of a large technical partner with resources in network and computing
  2711. infrastructure exceeding one&apos;s own, the desire to leverage geographically
  2712. diverse, redundant datacentres, the desire to fund computing from
  2713. operating rather than capital budgets, and the desire to build adaptable,
  2714. scriptable services with better connectivity to the Internet
  2715. than one could otherwise provide for oneself. But in the last year
  2716. or two, as anxiety about cybersecurity increases, I&apos;ve been hearing
  2717. more and more people refer to cybersecurity as their primary reason
  2718. for their adoption of the public cloud. I&apos;m not so sure what I think
  2719. of this reasoning. I can understand why someone might want to pass
  2720. to a third party a task that makes them anxious. In
  2721. situations involving strong emotions, such as anxiety, there is risk of &lt;A
  2722. HREF=&quot;https://www.psychologytoday.com/ca/blog/science-choice/201504/what-is-confirmation-bias&quot;&gt;
  2723. &quot;confirmation bias&quot;: believing something is true because you want it to be
  2724. true.&lt;/A&gt; But is it? &lt;A
  2725. HREF=&quot;https://www.investopedia.com/terms/c/ceterisparibus.asp&quot;&gt;Ceteris
  2726. paribus (all other things being equal)&lt;/A&gt;, is the public cloud intrinsicly
  2727. more secure than on-premise datacentres?
  2728.  
  2729. &lt;P&gt;
  2730. Some argue yes. Eplexity calls cloud computing &lt;A
  2731. HREF=&quot;https://eplexity.com/why-the-public-cloud-is-more-secure-than-an-on-premises-data-center/&quot;&gt;&quot;an
  2732. established best practice for businesses&quot; and claims
  2733. &quot;your data is typically safer in the public cloud than in an
  2734. on-premises data centre&quot;.&lt;/A&gt;
  2735. In 2016, Sara
  2736. Patrick of Clutch, guest-writing for Tripwire.com, claimed to have &lt;A
  2737. HREF=&quot;https://www.tripwire.com/state-of-security/security-data-protection/4-reasons-why-the-cloud-is-more-secure-than-legacy-systems/&quot;&gt;&quot;four
  2738. reasons why the Cloud is more secure than Legacy Systems&quot;&lt;/A&gt;
  2739. In 2017,
  2740. Quentin Hardy of the New York Times claimed that cloud data is &lt;A
  2741. HREF=&quot;https://www.nytimes.com/2017/01/23/insider/where-does-cloud-storage-really-reside-and-is-it-secure.html&quot;&gt;
  2742. &quot;probably more secure than conventionally stored data.&quot;&lt;/A&gt;
  2743. In 2018, David Linthicum, writing for InfoWorld, claimed &lt;A
  2744. HREF=&quot;https://www.infoworld.com/article/3316637/how-the-cloud-has-made-you-more-secure.html&quot;&gt;
  2745. &quot;your information is actually safer in the cloud than it is in your own data centre&quot;&lt;/A&gt;.
  2746.  
  2747. &lt;P&gt; One reason given for the claim is that public cloud providers offer greater
  2748. technical expertise than what is possible on-premise. Eplexity writes:
  2749. &lt;A HREF=&quot;https://eplexity.com/why-the-public-cloud-is-more-secure-than-an-on-premises-data-center/&quot;&gt;
  2750. &lt;blockquote&gt;
  2751. Unless your company is already in the business of IT security,
  2752. spending time and effort on securing your on-premises data distracts
  2753. from your core functions. Most organizations likely don&apos;t have a
  2754. robust, experienced team of cybersecurity professionals at their
  2755. disposal to properly protect their on-premises data.
  2756. ... As such, cloud providers may employ hundreds or thousands of
  2757. developers and IT professionals.
  2758. &lt;/blockquote&gt; &lt;/A&gt;
  2759.  
  2760. This is an argument from size and scale. Cloud providers are bigger than you,
  2761. and have arguably more IT expertise than you, so they can do a better job
  2762. than you. But sadly, size and IT expertise is no guarantee of security. &lt;A
  2763. HREF=&quot;https://qz.com/741056/the-stunning-collapse-of-yahoos-valuation/&quot;&gt;Yahoo
  2764. was a large Internet company, valued at one time at $125 billion.&lt;/A&gt;
  2765. It employed thousands of developers and IT professionals. Yet it was &lt;A
  2766. HREF=&quot;https://www.nytimes.com/2017/10/03/technology/yahoo-hack-3-billion-users.html&quot;&gt;subject
  2767. to a cybersecurity breach of three billion user accounts in 2013/14; the
  2768. breach was not disclosed until the fall of 2016, and the full impact was
  2769. not known until october 2017. The damage to Yahoo&apos;s business was significant:
  2770. Verizon acquired Yahoo in 2017 for less than $5 billion, a deal that was
  2771. nearly derailed by the disclosure of the breaches.&lt;/A&gt;
  2772.  
  2773. &lt;P&gt;I think we must conclude from the Yahoo story that size and expertise
  2774. alone is no guarantee of cybersecurity. Naturally, major cloud providers
  2775. like Amazon, Microsoft and Google are aware of the Yahoo situation and
  2776. its consequences. No doubt it illustrated for them the negative impact
  2777. that a major breach would have on their business. I cannot imagine that
  2778. they would take the threat lightly.
  2779.  
  2780. &lt;P&gt;Yet there have been close calls. Microsoft, a major cloud provider, &lt;A
  2781. HREF=&quot;https://www.zdnet.com/article/microsoft-discloses-security-breach-of-customer-support-database/&quot;&gt;
  2782. in December 2019 accidentally disclosed to the world a cloud database
  2783. on Azure with 250 million entries of customer support data. Happily,
  2784. a security researcher spotted and reported it, and Microsoft fixed it soon after.&lt;/A&gt; Moreover, &lt;A
  2785. HREF=&quot;https://www.forbes.com/sites/zakdoffman/2020/01/30/severe-perfect-100-microsoft-flaw-confirmed-this-is-a-cloud-security-nightmare/#7b5493dfb4a4&quot;&gt;Zak
  2786. Doffman, writing for Forbes, reported in Jan 2020 that Check Point Software
  2787. Technologies, a cybersecurity vendor, had discovered in 2019 a serious flaw
  2788. in Microsoft Azure&apos;s infrastructure that allowed users of the service to
  2789. access other users&apos; data.&lt;/A&gt; While Check Point reported it immediately to
  2790. Microsoft, who fixed it quickly, had the flaw been discovered by criminals
  2791. instead of cybersecurity researchers, a great many things running on Azure
  2792. could have been compromised. Doffman quotes Yaniv Balmas of Check Point: &lt;A
  2793. HREF=&quot;https://www.forbes.com/sites/zakdoffman/2020/01/30/severe-perfect-100-microsoft-flaw-confirmed-this-is-a-cloud-security-nightmare/#7b5493dfb4a4&quot;&gt;
  2794. &lt;blockquote&gt;
  2795. ...the take away here is that the big cloud concept of security
  2796. free from vulnerabilities is wrong. That&apos;s what we showed. It can
  2797. happen there as well. It&apos;s just software and software has bugs. The
  2798. fact I can then control the infrastructure gives me unlimited power.
  2799. &lt;/blockquote&gt;&lt;/A&gt; In the Check Point research article describing the flaw, Balmas concludes:
  2800.  
  2801. &lt;A HREF=&quot;https://research.checkpoint.com/2020/remote-cloud-execution-critical-vulnerabilities-in-azure-cloud-infrastructure-part-ii/&quot;&gt;
  2802. &lt;blockquote&gt;
  2803. The cloud is not a magical place. Although it is considered safe, it
  2804. is ultimately an infrastructure that consists of code that can have
  2805. vulnerabilities - just as we demonstrated in this article.
  2806. &lt;/blockquote&gt;
  2807. &lt;/A&gt;
  2808.  
  2809. &lt;P&gt;What, then, is the right answer? Well, there isn&apos;t one. Neither public
  2810. cloud or on-premise datacentres are magic, neither are &quot;safe&quot;. Cybersecurity
  2811. is a challenge that has to be met, no matter where the service is, or what
  2812. infrastructure it is using. Happily, this is finally being
  2813. recognized. Even Gartner Research, a long-time proponent
  2814. of the public cloud, predicting
  2815. as recently as mid-2019 that &lt;A
  2816. HREF=&quot;https://web.archive.org/web/20190808014445/https://www.gartner.com/smarterwithgartner/is-the-cloud-secure/&quot;&gt;public
  2817. cloud infrastructure as a service (IaaS) workloads will suffer at least
  2818. 60% fewer security incidents than those in traditional data centers&lt;/A&gt;, has recently taken a more nuanced view.
  2819. In the fall of 2019, this prediction of fewer security incidents in the cloud disappeared from Gartner&apos;s website,
  2820. and was replaced by this:
  2821. &lt;A HREF=&quot;https://web.archive.org/web/20191130104612/www.gartner.com/smarterwithgartner/is-the-cloud-secure/&quot;&gt;
  2822. &lt;blockquote&gt; &lt;B&gt;Through 2024, the majority of
  2823. enterprises will continue to struggle with appropriately measuring cloud
  2824. security risks.&lt;/B&gt; &lt;BR&gt;Questions around the security of public cloud
  2825. services are valid, but overestimating cloud risks can result in missed
  2826. opportunities. Yet, while enterprises tended to overestimate cloud risk
  2827. in the past, there&apos;s been a recent shift - many organizations are now
  2828. underestimating cloud risks. This can prove just as detrimental, if not
  2829. more so, than an overestimation of risk. A well-designed risk management
  2830. strategy, aligned with the overarching cloud strategy, can help organizations
  2831. determine where public cloud use makes sense and what actions can be taken
  2832. to reduce risk exposure. &lt;/blockquote&gt; &lt;/A&gt;
  2833. &lt;P&gt; So does &quot;public cloud use make sense&quot;? Yes, of course it does, for a great
  2834. many things.  But it&apos;s not because the public cloud is intrinsicly more secure. The
  2835. public cloud has its own set of cybersecurity issues. There is no &quot;free pass&quot;.
  2836. As always, carefully assess your risks and make an informed decision.
  2837. </description>
  2838.  </item>
  2839.  <item>
  2840.    <title>Does AI Help or Hinder Cybersecurity?</title>
  2841.    <pubDate>Fri, 24 Jan 2020 20:02:00 -0500</pubDate>
  2842.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525257Ejdd/blog.cgi/2020/01/24#aisecurity</link>
  2843.    <category>/it</category>
  2844.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525257Ejdd/blog.cgi/it/aisecurity</guid>
  2845.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  2846. &lt;IMG src=&quot;/~jdd/blog/it/security-4700815_640.jpg&quot; alt=&quot;Hooded figure with glowing circuit-board visage&quot;&gt;
  2847. &lt;figcaption&gt;
  2848. Image by &lt;a href=&quot;https://pixabay.com/users/geralt-9301/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=4700815&quot;&gt;Gerd Altmann&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=4700815&quot;&gt;Pixabay&lt;/a&gt;
  2849. &lt;/figcaption&gt;
  2850. &lt;/figure&gt;
  2851.  
  2852. Both AI and cybersecurity have become increasingly prominent in recent
  2853. years. AI&apos;s prominence has been driven by advances in machine learning and
  2854. the very real improvements it has made in the ability of computer
  2855. systems to do things that previously seemed possible only to human
  2856. beings. Cybersecurity&apos;s prominence has been driven by a number of
  2857. developments, including increasing nation-state conflict on the Internet,
  2858. and a dramatic rise in organized cyber-crime. It is inevitable that the two
  2859. will combine: AI will be and is being applied to the cybersecurity space,
  2860. through the development of machine learning techniques for breaking into
  2861. and defending systems.
  2862.  
  2863. &lt;P&gt; One view on this is that machine learning, as a powerful technique
  2864. that enables computer systems to take on tasks previously reserved only
  2865. for humans, will empower cyberattackers to breach computer security in
  2866. new ways, or at least in ways more effective than before. I know there is
  2867. a great deal of anxiety about this. This past fall, I had a conversation with
  2868. a CIO of a large university, who told me that his university was migrating
  2869. its internet services to Amazon precisely because he believed that new
  2870. AI-powered cyberattacks were coming, and he thought Amazon would be better
  2871. able to fend them off. I&apos;m not sure what I think of this defensive strategy,
  2872. but that is not the important question here. The key question is this:
  2873. are AI-powered cyberattacks going to overwhelm cyberdefence?
  2874.  
  2875. &lt;P&gt; No doubt AI-powered cyberattacks are a reality.  Machine learning is a
  2876. powerful computer science technique, especially for automation.
  2877. Cyberattackers, especially sophisticated, well-funded cyberattackers,
  2878. will use it and I am confident are already using it. But highly
  2879. automated cyberattacks are nothing new: cyberattackers have been automating
  2880. their attacks for decades. Smarter automated cyberattacks are certainly
  2881. something to worry about, but will they be transformative? Maybe. After
  2882. all, in cybersecurity, the advantage is to the attacker, who needs to find
  2883. only one hole in the defences, while the defender needs to block all of them.
  2884. Anything that boosts the effectiveness of the attacker would seem to
  2885. make the situation worse.
  2886.  
  2887. &lt;P&gt; To really see the full picture, it&apos;s important to look at the
  2888. defender too.  Machine learning makes the situation worse only if it
  2889. benefits the attacker more than it benefits the defender.  But does it?
  2890.  
  2891. &lt;P&gt; I don&apos;t have a complete answer to this question: there is a great
  2892. deal of work still to be done on the application of machine learning
  2893. to cybersecurity. But I suspect that the answer is a qualified &lt;em&gt;No&lt;/em&gt;:
  2894. rather, all other things being equal, machine learning will likely shift
  2895. the balance of power towards the defender. The reason is data.
  2896.  
  2897. &lt;P&gt; Machine learning is a technique where computer systems, instead of being
  2898. programmed by programmers, learn what to do from data.  But the quality of
  2899. the learning depends on the quality and in particular the quantity of data.
  2900. Machine learning is a technique that is most effective when trained with
  2901. large amounts of data. &lt;A HREF=&quot;http://www.image-net.org&quot;&gt;ImageNet&lt;/A&gt;,
  2902. for instance, a standard training dataset used to train machine learning
  2903. applications to recognize images, contains about 14.2 million images.
  2904. But who is more likely to have access to large amounts of good data
  2905. about a system: the attacker or the defender?  Of course, it depends, but
  2906. it seems to me that, very generally speaking, the defender is more likely
  2907. to have access to good system data than the attacker. The attacker is
  2908. trying to get in; the defender is already in.
  2909.  
  2910. &lt;P&gt; Of course, this is the broadest of generalizations. The effectiveness
  2911. of machine learning in the cybersecurity space depends on a great many
  2912. things. But I am cautiously optimistic. I realize I may be bucking what
  2913. seems to be becoming a prevailing trend of ever-increasing anxiety about
  2914. cybersecurity, but I believe here that machine learning has more potential
  2915. to help than to harm. I look forward to seeing what will emerge in this
  2916. space over the next few years.
  2917. </description>
  2918.  </item>
  2919.  <item>
  2920.    <title>What&apos;s all the fuss about AI anyway?</title>
  2921.    <pubDate>Mon, 30 Sep 2019 00:00:00 -0400</pubDate>
  2922.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525257Ejdd/blog.cgi/2019/09/30#ai</link>
  2923.    <category>/it</category>
  2924.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525257Ejdd/blog.cgi/it/ai</guid>
  2925.    <description>&lt;figure style=&quot;float:left;width:50%;padding:1px&quot;&gt;
  2926. &lt;IMG src=&quot;/~jdd/blog/it/images/a-2729794_640.png&quot; alt=&quot;Brain-shaped Network&quot;&gt;
  2927. &lt;figcaption&gt;
  2928. Image by &lt;a href=&quot;https://pixabay.com/users/GDJ-1086657/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=2729794&quot;&gt;Gordon Johnson&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=2729794&quot;&gt;Pixabay&lt;/a&gt;
  2929. &lt;/figcaption&gt;
  2930. &lt;/figure&gt;
  2931.  
  2932. A great deal in the past five years has been written about Artificial
  2933. Intelligence (AI). But there&apos;s a lot of confusion about what AI actually
  2934. is, and why it is of special interest now.  Lets clear up some of that
  2935. confusion. In ordinary language, what is this fuss about AI all about?
  2936.  
  2937. &lt;P&gt;
  2938. AI, broadly understood, is a term used to describe a set of computing
  2939. techniques that allow computers to do things that human beings use
  2940. intelligence to do. This is not to say that the computer is intelligent, but
  2941. rather that the computer is doing something that, if done by a person, would
  2942. be considered evidence of that person&apos;s intelligence. Contrary to widespread
  2943. opinion, this is not the same thing as an artificial person.  In fact, there
  2944. have been for a long time many things that humans use intelligence to do,
  2945. that computers do better, whether it be remembering and
  2946. recalling items, doing arithmetic, or playing chess. But computers
  2947. do these things using different techniques than humans do. For example, &lt;A
  2948. HREF=&quot;https://www.chess.com/article/view/deep-blue-kasparov-chess&quot;&gt;Deep
  2949. Blue, a custom chess computer built by IBM, beat Garry Kasparov, the
  2950. then-reigning world chess champion, in 1997&lt;/A&gt;, but Deep Blue played chess
  2951. in a very different way than Garry. Garry relied on his human intelligence,
  2952. while Deep Blue used programming and data.
  2953.  
  2954. &lt;P&gt;
  2955. However, some computer scientists, noting that people can do things that
  2956. computers can&apos;t, thought long and hard about ways that people do it,
  2957. and how computers might be progammed to do the same. One such technique,
  2958. deep learning, a neural network technique modelled after the human brain,
  2959. has been worked on since the 1980s, with slow but steady improvement,
  2960. but computer power was limited and error rates were often high, and for
  2961. many years, most computer scientists seemed to feel that other techniques
  2962. would yield better results. But a few kept at it, knowing that the computers
  2963. of the day were inadequate, but advances in computing would make things
  2964. possible that weren&apos;t possible before.
  2965.  
  2966. &lt;P&gt;
  2967. This all changed in 2012, when one such researcher,
  2968. Geoff Hinton, and his students, working here at the University of Toronto, published a &lt;A
  2969. HREF=&quot;https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf&quot;&gt;seminal
  2970. deep learning paper that cut error rates dramatically.&lt;/A&gt;
  2971. I remember supporting Geoff&apos;s group&apos;s research computing at that time.
  2972. It was a bit challenging: we were using multiple GPUs per machine to
  2973. train machine learning models at a time when GPU computing was still rather
  2974. new and somewhat unreliable. But GPUs were absolutely necessary: without
  2975. them, instead of days of computing time to train a model, months would be
  2976. required.  One of our staff, Relu Patrascu, a computer scientist and skilled
  2977. system administrator working hand-in-glove with the researchers, tuned and
  2978. configured and babysat those machines as if they were sick children. But it
  2979. worked! Suddenly deep learning could produce results closer to what people
  2980. could do, and that was only the beginning. Since then, deep learning has
  2981. produced terrific results in all sorts of domains, some exceeding what
  2982. people can do, and we&apos;ve not even scraped the surface of what is possible.
  2983.  
  2984. &lt;P&gt; But what does deep learning actually do?  It is a computer
  2985. science data classification technique. It&apos;s used to take input data and
  2986. classify it: give it a thing and it will figure out what the thing is.
  2987. But it classifies things in a way that&apos;s different and more useful than
  2988. traditional computer science methods for classification, such as computer
  2989. programming, or data storage and retrieval (databases). As such, it can be
  2990. used to do a lot more than computers previously had been able to do.  &lt;P&gt;
  2991. To see this, consider traditional computer science methods: for example,
  2992. computer programming. This approach requires a person to write code that
  2993. explicitly considers different cases. For example, imagine that you want to
  2994. classify two-dimensional figures.  You want to consider whether they are
  2995. regular polygons. You could write a computer program that defines for itself
  2996. what a regular polygon is, and checks each characteristic of an input shape
  2997. to see whether or not it matches the definition of a regular polygon. Such a
  2998. program, when given a square, will notice that it is a polygon, it has four
  2999. sides, and that those sides are equal in length. Since the programmer
  3000. put into the program a detailed definition of what a regular polygon is,
  3001. and since the program checks each feature explicitly, it can tell whether
  3002. or not a shape is a regular polygon, even if the program has never seen
  3003. that particular shape before.
  3004.  
  3005. &lt;P&gt;But what about exceptional cases? Is a circle a regular polygon? It is,
  3006. after all, the limit of an N-gon as N goes to infinity. This is an &quot;edge
  3007. case&quot; and programs need to consider those explicitly. A programmer had to
  3008. anticipate this case and write it into the program. Moreover, if you wanted to
  3009. consider some other type of shape, a programmer would have to rewrite
  3010. the code accordingly. There&apos;s no going from a bunch of examples to working
  3011. code without a programmer to write it.  Programming is certainly a useful
  3012. technique, but it has its limits. Wouldn&apos;t it be nice to be able to
  3013. learn from a bunch of examples, without a person having to write all that
  3014. code?
  3015.  
  3016. &lt;P&gt; One way to do that would be data storage and retrieval, for example,
  3017. a database.  Consider the shape classifier problem again. You might put
  3018. in a bunch of shapes into a database, indicating whether the shape is
  3019. a regular polygon or not. Once the database is populated, classifying a
  3020. shape simply becomes looking it up. The database will say whether or not
  3021. it is a regular polygon.
  3022.  
  3023. &lt;P&gt; But what if it&apos;s not there? A database has the advantage of being
  3024. able to learn from examples. But it has a big disadvantage: if it hasn&apos;t
  3025. seen an example before, and is asked about it, it has no idea what the
  3026. right answer is. So while data storage and retrieval is a very useful
  3027. computing technique, and it is the backbone of most of our modern
  3028. information systems, it has its limits. Wouldn&apos;t it be nice if
  3029. a classifier system could provide a useful answer for input data
  3030. that it&apos;s never seen before, without a programmer to tell it how?
  3031.  
  3032. &lt;P&gt; Deep learning does exactly this. Like data storage and retrieval, it
  3033. learns from examples, through training. Very roughly, a neural network,
  3034. when trained, is given some input data, and is told what output
  3035. data it should produce when it sees that data in future. These input and
  3036. output constraints propagate forward and backwards through the network,
  3037. and are used to modify internal values such that when the network next
  3038. sees input like that, it will produce the matching output.
  3039.  
  3040. &lt;P&gt; The key advantage of this technique is that if it sees data that
  3041. is similar to, but not the same as data it has been trained on, it will
  3042. produce output similar to the trained output. This is very important,
  3043. because like programming, it can work on input it has never seen, but like
  3044. databases, it can learn from examples and need not be coded by a programmer
  3045. anticipating all the details in advance. For our shape example, if trained
  3046. with many examples of regular polygons, the neural network will be able to
  3047. figure out whether or not a given input is a regular polygon, and perhaps
  3048. even more interestingly, it will be able to note that a circle is very
  3049. like a regular polygon, even if it had never been trained on a circle.
  3050.  
  3051. &lt;P&gt; Moreover, a deep learning neural network can learn from its own
  3052. results. This is called reinforcement learning. This technique involves
  3053. using a neural network to derive output data from some input data, the
  3054. results are tested to see how well they work, and the neural network is
  3055. retrained accordingly.  This way a neural network can &quot;learn from its own
  3056. mistakes&quot;, training itself iteratively to classify better. For example, a
  3057. model of a walking human, with some simple programming to teach it the laws
  3058. of physics, can, using reinforcement learning, teach itself how to walk. A
  3059. few years ago, some of the researchers in our department
  3060. did exactly that.  Another example: Google got a lot of
  3061. attention a few years ago when deep learning researchers there &lt;A
  3062. HREF=&quot;https://deepmind.com/research/case-studies/alphago-the-story-so-far&quot;&gt;
  3063. built a deep learning system that used reinforcement learning to become a
  3064. champion at the game of Go, a game very hard to computerize using traditional
  3065. techniques, and proved it by beating the reigning Go world champion. &lt;/A&gt;
  3066.  
  3067.  
  3068. &lt;P&gt; It seems clear to me at this point that deep learning is as fundamental
  3069. a computing technique as computer programming and databases in building
  3070. practical computer systems. It is enormously powerful, and is causing a
  3071. great deal of legitimate excitement. Like all computer science techniques, it
  3072. has its advantages and drawbacks, but its strengths are where other computer
  3073. science techniques have weaknesses, and so it is changing computer science
  3074. (and data science more generally) in dramatic ways. It&apos;s an interesting time
  3075. to be a computer scientist, and I can&apos;t even begin to imagine the many things
  3076. that bright and innovative people will be able to do with it in the future.
  3077. </description>
  3078.  </item>
  3079.  <item>
  3080.    <title>Existential threats from AI?</title>
  3081.    <pubDate>Mon, 02 Sep 2019 20:14:00 -0400</pubDate>
  3082.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525257Ejdd/blog.cgi/2019/09/02#aithreats</link>
  3083.    <category>/it</category>
  3084.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525257Ejdd/blog.cgi/it/aithreats</guid>
  3085.    <description>&lt;figure style=&quot;float:left;padding:1px&quot;&gt;
  3086. &lt;IMG src=&quot;/~jdd/blog/it/nuclear-2136244_640.jpg&quot; alt=&quot;Nuclear explosion&quot;&gt;
  3087. &lt;figcaption&gt;
  3088. Image by &lt;a href=&quot;https://pixabay.com/users/AlexAntropov86-2691829/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=2136244&quot;&gt;Alexander Antropov&lt;/a&gt; from &lt;a href=&quot;https://pixabay.com/?utm_source=link-attribution&amp;amp;utm_medium=referral&amp;amp;utm_campaign=image&amp;amp;utm_content=2136244&quot;&gt;Pixabay&lt;/a&gt;
  3089. &lt;/figcaption&gt;
  3090. &lt;/figure&gt;
  3091.  
  3092. Plenty has been written about the possible threats to humanity from Artificial
  3093. Intelligence (AI). This is an old concern, a staple of science fiction since
  3094. at least the 1950s. The usual story: a machine achieves
  3095. sentience and pursues its own agenda, harmful to people. The current successes
  3096. of machine learning have revived this idea. The &lt;A HREF=&quot;https://www.bbc.com/news/technology-30290540&quot;&gt;late Stephen
  3097. Hawking warned the BBC in 2014&lt;/A&gt; that &quot;the development of full artificial
  3098. intelligence could spell the end of the human race&quot;.  He feared that &quot;it
  3099. would take off on its own, and re-design itself at an ever increasing rate.&quot;
  3100. He worries that human beings, &quot;who are limited by slow biological evolution, couldn&apos;t compete, and would be superseded.&quot;
  3101. &lt;A
  3102. HREF=&quot;https://www.theatlantic.com/magazine/archive/2018/06/henry-kissinger-ai-could-mean-the-end-of-human-history/559124/&quot;&gt;Henry
  3103. Kissinger, in a thoughtful essay in &lt;em&gt;The Atlantic&lt;/em&gt; last
  3104. year&lt;/A&gt;, worried that &quot;AI, by mastering certain competencies more
  3105. rapidly and definitively than humans, could over time diminish human
  3106. competence and the human condition itself as it turns it into data.&quot;  &lt;A
  3107. HREF=&quot;https://www.wired.com/story/elon-musk-humanity-biological-boot-loader-ai/&quot;&gt;Elon
  3108. Musk, in a debate last month with Alibaba&apos;s Jack Ma, reported by WIRED,
  3109. &lt;/A&gt; argued that &quot;there&apos;s just a smaller and smaller corner of
  3110. what of intellectual pursuits that humans are better than computers. And
  3111. that every year, it gets smaller and smaller, and soon will be far far
  3112. surpassed in every single way. Guaranteed. Or civilization will end.&quot;
  3113.  
  3114. &lt;P&gt;Are they right? Is there an existential threat to humanity from
  3115. AI? Well, yes, I think there actually is one, but not quite in the way
  3116. Musk, Kissinger, or Hawking fear. Computer have been better at humans for
  3117. a long time in many cognitive domains. Computers remember
  3118. things more accurately, process things faster, and scale better
  3119. than humans in many tasks. AI, particularly machine learning, increases
  3120. the number of skills where computers are better than humans. Given that
  3121. humanity has been spending the last couple of generations getting used to
  3122. a certain arrangement where computers are good at some things and humans
  3123. are good at others, it can be a bit disconcerting to have this upended by
  3124. computers suddenly getting good at things they weren&apos;t good at before. I
  3125. understand how this can make some people feel insecure, especially highly
  3126. accomplished people who define themselves by their skills. Kissinger,
  3127. Musk and Hawking fear a world in which computers are better at many things
  3128. than humans. But we have been living in such a world for decades. AI simply
  3129. broadens the set of skills in question.
  3130.  
  3131. &lt;P&gt;As a computer scientist, I am not particularly worried about the notion of
  3132. computers replacing people. Yes, computers are developing new useful skills,
  3133. and it will take some getting used to. But I see no imminent danger of
  3134. AI resulting in an artificial person, and even if it did, I don&apos;t think
  3135. an artificial person is an intrinsic danger to humans.  Yet I agree that
  3136. there are real existential threats to humanity posed by AI. But these are
  3137. not so much long term or philosophical, to me they&apos;re eminently practical
  3138. and immediate.
  3139.  
  3140. &lt;P&gt; The first threat is the same sort of threat as posed by nuclear
  3141. physics: AI can be used to create weapons that can cause harm to people on a massive scale.  Unlike nuclear bombs, AI weapons do not do their harm through
  3142. sheer energy discharge. Rather, machine learning, coupled with advances in
  3143. miniaturization and mass production, can be used to create horrific smart
  3144. weapons that learn, swarms of lethal adaptive drones that seek out and
  3145. destroy people relentlessly. A deep commitment to social responsibility,
  3146. plus a healthy respect for the implications of such weapons, will be needed
  3147. to offset this danger.
  3148.  
  3149. &lt;P&gt; The second threat, perhaps even more serious, comes not from AI itself
  3150. but from the perceptions it creates. AI&apos;s successes are transforming
  3151. human work: because of machine learning, more and more jobs,
  3152. even white-collar ones requiring substantial training, can be replaced by
  3153. computers. It&apos;s unclear yet to what extent jobs eliminated by AI will be offset
  3154. by new jobs created by AI, but if AI results in a widespread perception
  3155. that most human workers are no longer needed, this perception may itself
  3156. become an existential threat to humanity. The increasingly obvious fact
  3157. of anthropogenic climate change has already fueled the idea that humanity
  3158. itself can be viewed as an existential threat to the planet. If AI makes
  3159. it possible for some to think that they can have the benefits of society
  3160. without keeping many people around to do the work, I worry we may see
  3161. serious consideration of ways to reduce the human population to much
  3162. smaller numbers. This to me is a dangerous and deeply troubling idea,
  3163. and I believe a genuine appreciation for the intrinsic value of all human
  3164. beings, not just those who are useful at the moment, will be needed to
  3165. forestall it. Moreover, a good argument from future utility can also be
  3166. made: we cannot accurately predict which humans will be the great inventors
  3167. and major contributors of the future, the very people we need to address
  3168. anthropogenic climate change and many other challenges. If we value all
  3169. people, and build a social environment in which everyone can flourish,
  3170. many innovators of the future will emerge, even from unexpected quarters.
  3171.  
  3172. &lt;P&gt;Threats notwithstanding, I don&apos;t think AI or machine learning
  3173. can go back into Pandora&apos;s box, and as a computer scientist who has been
  3174. providing computing support for machine learning since long before it
  3175. became popular, I would not want it to. AI is a powerful tool, and like
  3176. all powerful tools, it can be used for many good things. Let us build a
  3177. world together in which it is used for good, not harm.
  3178. </description>
  3179.  </item>
  3180.  <item>
  3181.    <title>Why we thought for a while Pluto was a planet, but it never was.</title>
  3182.    <pubDate>Mon, 26 Aug 2019 06:51:00 -0400</pubDate>
  3183.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525257Ejdd/blog.cgi/2019/08/26#pluto</link>
  3184.    <category>/misc</category>
  3185.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525257Ejdd/blog.cgi/misc/pluto</guid>
  3186.    <description>&lt;IMG src=&quot;/~jdd/blog/misc/pluto-1315109_640.jpg&quot; alt=&quot;Pluto&quot; style=&quot;float:left;padding:10px&quot;&gt;
  3187.  
  3188. &lt;P&gt; More than a decade after Pluto&apos;s demotion from the rank of planet, some
  3189. still do not accept it. I can sympathize. Like many of us, I grew up
  3190. memorizing in school the nine planets of the Solar system, the last of
  3191. which was Pluto: icy, distant and mysterious. I remember as a child poring
  3192. over a diagram of the solar system, marvelling at the concentric elipses of
  3193. the planetary orbits, and wondering why Pluto&apos;s orbit was so odd. For odd
  3194. it was: all the other planets orbited the sun in more or less concentric
  3195. elipses, but Pluto was eccentric: its orbit was at an unusual angle, and
  3196. it even briefly came closer to the sun than Neptune.  None of the other
  3197. plants had orbits like this: why Pluto? But I didn&apos;t question that it was a
  3198. planet. It had been recognized as a planet since Clyde Tombaugh discovered
  3199. it before my parents were born. For me, Pluto was weird, but it was still
  3200. &quot;planet&quot;, the astronomical equivalent of a sort of odd uncle who behaved
  3201. strangely and kept to himself, but still family.
  3202.  
  3203. &lt;P&gt; But the idea of Pluto as a planet started to become problematic in
  3204. the early 1990s. In 1992, Jewitt and Luu discovered another object beyond
  3205. Neptune: Albion, much smaller than Pluto, and also with an odd orbit. Because
  3206. it was a small object, it was pretty clearly not a planet, so Pluto&apos;s
  3207. status was not yet in question, but it was only the first of many. By 2000,
  3208. more than seventy such objects had been discovered. Most of these were very
  3209. small, but some were not so small. And the discoveries continued. In 2003,
  3210. with the discovery of the Eris, a trans-Neptunian body more
  3211. massive than Pluto itself, the problem became acute. No longer was Pluto
  3212. the odd uncle of the planets: now there were on the order of 100 odd
  3213. uncles and aunts, and at least one of them, Eris, aptly named after the greek
  3214. goddess of discord, had a better claim to planethood than Pluto itself.
  3215. Something had to be done.  This bunch of odd objects, odd in the same way as
  3216. Pluto, were either all planets, or they were none of them planets. There
  3217. was no reasonable distinction that could be made that would keep Pluto
  3218. a planet but deny planethood to Eris and many of her siblings. To do so
  3219. would be arbitrary: we would be saying that Pluto was a planet simply because
  3220. we discovered it first and it took us a long time to discover the others.
  3221. What to do?
  3222.  
  3223. &lt;P&gt; Happily, there was a precedent: this sort of thing had come up
  3224. before. In 1801, Giuseppe Piazza discovered Ceres,
  3225. a body orbiting between Mars and Jupiter. This was a big deal.  Only twenty
  3226. years before, a new planet had been discovered for the first time in
  3227. recorded history: Uranus, found by accident by William Herschel in 1781. Now,
  3228. twenty years later, Piazza had found a second.  And this one was not out beyond
  3229. Saturn, it was nearer than Jupiter. But Piazza&apos;s share of the limelight
  3230. was soon to lessen. his planet had a rival: a year later, Heinrich Wilhelm
  3231. Olbers discovered Pallas, another body between Jupiter and Mars. Two
  3232. years later, in 1804, Karl Harding discovered another: Juno. Not to be
  3233. outdone, Olbers in 1807 discovered yet another, Vesta.  By the middle of
  3234. the 19th century, fifteen bodies orbiting between Mars and Jupiter
  3235. were known, and while none of them were anywhere as large as Ceres, one
  3236. of them, Vesta, had nearly a third of Ceres&apos; mass. Were there really many
  3237. small planets between Mars and Jupiter, or were these something else?
  3238. When in 1846 the planet Neptune was discovered beyond Uranus, it became
  3239. clear that some decision about these bodies between Mars and Jupiter needed
  3240. to be made. A consensus emerged: Ceres and other such objects were not
  3241. planets. They were called &quot;asteroids&quot;, a name coined in 1802 by William
  3242. Herschel. It was a good call: there are now well over 100,000 known
  3243. asteroids, far too many for schoolchildren to memorize.
  3244.  
  3245. &lt;P&gt; With Pluto, a similar situation was now occurring. While we weren&apos;t
  3246. yet at 100,000 Pluto-like bodies, we knew about quite a few more than
  3247. fifteen. And Pluto, unlike Ceres, wasn&apos;t even the most massive: Eris was,
  3248. and quite possibly, bigger ones would be found. There was no denying the
  3249. facts. Pluto, like Ceres, could not be a planet. It must be something else.
  3250.  
  3251. &lt;P&gt; Of course this was quite controversial. People had been calling Pluto
  3252. a planet for the better part of a century. Generations of schoolchildren
  3253. had memorized it as part of the list of planets. But the choice was clear:
  3254. either the schoolchildren would have to start memorizing longer lists,
  3255. much much longer ones, or Pluto would have to be demoted. Well, not
  3256. demoted, exactly, but newly recognized for what it really was all
  3257. along: something different. In the sumer of 2006, the &lt;A
  3258. HREF=&quot;http://www.iau.org&quot;&gt;International Astronomical Union
  3259. (IAU)&lt;/A&gt; declared that Pluto isn&apos;t a planet, it is a &lt;A
  3260. HREF=&quot;https://www.iau.org/static/resolutions/Resolution_GA26-5-6.pdf&quot;&gt;&lt;EM&gt;dwarf
  3261. planet&lt;/EM&gt;&lt;/A&gt;. While this designation is a little confusing (if a
  3262. dwarf planet isn&apos;t a planet, why is it called a dwarf &lt;EM&gt;planet&lt;/EM&gt;?),
  3263. one thing was now clear: Pluto is not the same sort of thing as Mercury,
  3264. Venus, Earth, Mars, Jupiter, Saturn, Uranus and Neptune; it, and Eris,
  3265. and probably a couple of other larger trans-Neptunian bodies discovered
  3266. since the 1990s, are something different.  But guess what: Ceres, too,
  3267. fits IAU&apos;s definition of dwarf planet, the only asteroid that does. Two
  3268. centuries after its discovery, Ceres, first-born of the non-planets
  3269. and largest of the asteroids, was deemed a dwarf planet, and Piazza, its
  3270. discoverer, though not the second person in recorded history to discover
  3271. a new planet, was recognized as the very first to discover a dwarf one.
  3272. </description>
  3273.  </item>
  3274.  <item>
  3275.    <title>Ross Anderson&apos;s &lt;EM&gt;Security Engineering&lt;/EM&gt;</title>
  3276.    <pubDate>Fri, 19 Jul 2019 16:13:00 -0400</pubDate>
  3277.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525257Ejdd/blog.cgi/2019/07/19#anderson</link>
  3278.    <category>/it</category>
  3279.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525257Ejdd/blog.cgi/it/anderson</guid>
  3280.    <description>&lt;A HREF=&quot;https://www.cl.cam.ac.uk/~rja14/book.html&quot;&gt;
  3281. &lt;IMG src=&quot;https://www.cl.cam.ac.uk/~rja14/Papers/book2coversmall.jpg&quot; alt=&quot;Security Engineering - Second Edition&quot; style=&quot;float:left;padding:10px&quot;&gt;&lt;/A&gt;
  3282.  
  3283. &lt;P&gt;Until recently, I had not read Ross Anderson&apos;s &lt;EM&gt;Security
  3284. Engineering&lt;/EM&gt;, despite hearing good things about it. I&apos;m not sure why:
  3285. I think I was put off a bit by the title. I had a vague and confused
  3286. impression that a book about &quot;Security Engineering&quot; would be yet another
  3287. how-to book about making computers secure. I should have known better. In
  3288. this case, I was wrong, very much so, and much to my detriment. I should
  3289. have read this book long ago.
  3290.  
  3291. &lt;P&gt; Why had I not read it?  I have no excuse. The book has
  3292. been out for a while: it is in its second edition, which came
  3293. out in 2008 (Anderson is writing a third edition, expected next
  3294. year). So I certainly had the opportunity.  Moreover, since 2012,
  3295. the book has been free for the reading (and downloading) from &lt;A
  3296. HREF=&quot;https://www.cl.cam.ac.uk/~rja14/book.html&quot;&gt;his website&lt;/A&gt;. So I
  3297. certainly had the means.  I just didn&apos;t, until a few weeks ago, when I
  3298. stumbled across it again. I read a little from the website, then a little
  3299. more. Before long, I was well and thoroughly hooked.
  3300.  
  3301. &lt;P&gt;&lt;EM&gt;Security Engineering&lt;/EM&gt; is a classic, comprehensive book about
  3302. information security: eminently readable, clear and thorough, it covers
  3303. information security in pretty much every aspect one might encounter it,
  3304. from the usual (cryptography, access controls, protocols, biometrics) to the
  3305. not quite so day-to-day (nuclear weapons launch protocols, counterfeiting,
  3306. even spying by analyzing the RF emissions from computers). Each chapter
  3307. is a clear elucidation of a particular aspect of information security,
  3308. focusing on the essential issues. Each chapter provides enough detail to
  3309. understand the essential elements, yet not too much detail as to overwhelm
  3310. the reader. His writing is a classic illustration of the difference between
  3311. an expert and a master. An expert knows a great deal about a topic and
  3312. provides an abundance of information. A master knows the key elements,
  3313. those things that are most important, on which everything else hangs, and
  3314. focuses exactly on these. This book is mastery, in clear, understandable
  3315. and engaging language. It has become my favourite book in information
  3316. security already, and I haven&apos;t yet finished it.
  3317.  
  3318. &lt;P&gt;I look forward to the third edition sometime next year. I can&apos;t wait.
  3319. </description>
  3320.  </item>
  3321.  <item>
  3322.    <title>Externality and Information Security</title>
  3323.    <pubDate>Mon, 04 Mar 2019 12:04:00 -0500</pubDate>
  3324.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525257Ejdd/blog.cgi/2019/03/04#externality</link>
  3325.    <category>/it</category>
  3326.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525257Ejdd/blog.cgi/it/externality</guid>
  3327.    <description>
  3328. It was a hot midsummer weekend, and I was traveling back to Toronto
  3329. with friends. We were on the expressway (the name here in Ontario for
  3330. the sort of road that Americans call freeways and Brits call motorways).
  3331. Traffic was very slow: a classic traffic jam. After about thirty minutes,
  3332. we reached the cause of the problem. It was not a collision. Nor was it
  3333. highway construction. Instead, by the side of the roadway, a minivan was
  3334. parked, back gate open, and a family was having a picnic on the nearby
  3335. grass. I don&apos;t know if they realized they were causing a traffic jam, but
  3336. they were. People had slowed to look, which caused traffic behind to slow
  3337. too, and because of the traffic volume, this led to a traffic jam over a
  3338. considerable distance.
  3339.  
  3340. &lt;P&gt; I don&apos;t know why the family having the picnic had chosen that spot for
  3341. it, and I don&apos;t know whether they realized the problem they were causing.
  3342. But their picnic went on, unaffected by the traffic problems they were
  3343. causing. In other words, the traffic jam was &lt;em&gt;not their problem&lt;/em&gt;.
  3344. It was an &lt;em&gt;externality&lt;/em&gt;, something causing a negative effect not
  3345. felt by those who cause it.
  3346.  
  3347. &lt;P&gt; Externalities happen in life all the time. Large organizations
  3348. (companies, countries, institutions) suffer significantly when their
  3349. decision-makers make decisions that are good for themselves but not good
  3350. for the organization. Rules to make this less likely are put in place:
  3351. rules against bribery, rules concerning conflict of interest, rules imposing
  3352. due process. But rules only work to a certain extent: there are plenty of
  3353. situations where the rules are followed yet still externalities happen.
  3354. Moreover, rules come with costs, sometimes significant ones.  Rules may
  3355. be necessary, but they are not sufficient, and they need to be accompanied
  3356. by buy-in.
  3357.  
  3358. &lt;P&gt;Let&apos;s consider traffic again. Driving is governed by all sorts of rules.
  3359. Some of these rules work well: at traffic lights, go when the light is
  3360. green, stop when it is red.  Rarely broken, this rule makes traffic work
  3361. in dense situations where otherwise there would be chaos. Most of the
  3362. time, this rule is followed even in the absence of external enforcement.
  3363. When enforcement does occur, it is well regarded: hardly anyone will argue
  3364. that a person running a red light is a safety hazard and should be ticketed.
  3365. In practice, you can stand for hours beside a busy traffic signal in a
  3366. typical Ontario city, and despite the absence of police presence, not
  3367. find a single driver running a red light.
  3368.  
  3369. &lt;P&gt;Sadly, other driving rules don&apos;t work quite so well, such as speed limits
  3370. on expressways here in Ontario. These limits are often broken, with some
  3371. following them and others not. Often, on an uncongested expressway,
  3372. unless enforcement is likely (i.e. police is present) there will be some
  3373. people driving over the speed limit. Enforcement is viewed cynically:
  3374. speeding tickets are often viewed more as revenue generation than as a
  3375. safety measure. Obeying speed limits is often viewed by drivers as
  3376. an externality: &lt;em&gt;not my problem&lt;/em&gt;, unless there is a police officer
  3377. around to make it one. In practice, at any place on any uncongested
  3378. Ontario expressway, you will be hard-pressed to find a five-minute period
  3379. in which no passing driver has exceeded the speed limit.
  3380.  
  3381. &lt;P&gt;I have been thinking a lot about information security lately. In
  3382. information security, we have a situation similar in many respects to
  3383. driving. Just as driving is a matter of traveling safely, information
  3384. security is a matter of computing safely. When we compute, we may
  3385. be processing information that is sensitive, confidential, private. Harm
  3386. can occur when it is exposed.  Steps need to be taken to ensure that it
  3387. is not: persons handling information will have to handle it securely. But
  3388. do we want this process to look like speed limits? Or traffic lights? I
  3389. think the answer is clear: if we want information to actually be secure,
  3390. we want good security practice to be followed like the rules for traffic
  3391. lights are followed: broadly and consistently, without the need for the
  3392. constant threat of enforcement.
  3393.  
  3394. &lt;P&gt;In recent years, an information security profession has
  3395. arisen.  The increasing demands of the profession have made it increasingly
  3396. rare that an information security professional has spent much time actually
  3397. running a substantial IT operation.  Certifications abound, and a
  3398. multiplicity of complex and large security standards have been created,
  3399. each requiring professionals to interpret.  A great deal of money is
  3400. being spent on information security. Much of this is good and necessary:
  3401. information security needs attention, codification, dissemination, and
  3402. championship.  But the professionalization of information security comes
  3403. with big risks, too: the risk that information security will become the
  3404. responsibility only of specialists, the risk that these specialists will
  3405. come up with all-encompassing codexes of security standards to impose,
  3406. the risk that these standards will be treated as externalities by IT
  3407. practitioners, the risk that the information security profession will
  3408. respond with enforcement, and hence the risk we will find ourselves in
  3409. the expressway speed limit situation with respect to information security.
  3410.  
  3411. &lt;P&gt;The fact is, information security is an aspect of good IT practice:
  3412. if an implementation is not secure, it is &lt;em&gt;broken&lt;/em&gt;, just as much
  3413. as if it were not reliable. Security is the responsibility of all IT
  3414. practitioners: it needs to be internalized, not externalized.
  3415.  
  3416. &lt;P&gt;For this to happen, it is important that information security rules
  3417. be simple and understandable, to ensure buy-in.  Just as traffic light
  3418. rules address the obvious risk of traffic accidents, so should security
  3419. rules address clear risks in a visibly appropriate way. In most cases,
  3420. it&apos;s not so important that rules be part of a comprehensive codex that
  3421. addresses all possible areas of risk: the more complex the rule and the
  3422. more extensive the system of rules, the more likely it will all be treated
  3423. as an externality. What we really want are not rules for their own sake,
  3424. but genuinely secure IT.
  3425.  
  3426. &lt;P&gt; If we want secure IT, we need to recognize that there is another
  3427. potential externality at work. Genuine information security and the good of
  3428. the information security profession may not always align. Just as expressway
  3429. speed limits employ more police than traffic lights, an enforcement approach
  3430. will employ more information security professionals than an internalized
  3431. one. But the internalized approach is what gives us secure computing. This
  3432. is not something that can be left to the information security profession
  3433. alone. To get there, we will need collaborative effort from all of us,
  3434. particularly those with long experience running substantial IT operations.
  3435. We will all need to make a true commitment to a practical approach, one
  3436. that seeks to make computing genuinely more secure in the real world.
  3437. </description>
  3438.  </item>
  3439.  <item>
  3440.    <title>Intentionality</title>
  3441.    <pubDate>Tue, 26 Feb 2019 06:27:00 -0500</pubDate>
  3442.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525257Ejdd/blog.cgi/2019/02/26#intentionality</link>
  3443.    <category>/misc</category>
  3444.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/intentionality</guid>
  3445.    <description>
  3446. &lt;p&gt; I spent all of 2018 intending to blog, and not doing it. Sadly, this
  3447. is an all too human situation. We intend to do things, when we can, when
  3448. time permits, but we can&apos;t; time doesn&apos;t permit. Or at least this is
  3449. one of those stories we tell ourselves. The truth is a little simpler:
  3450. throughout 2018, my intention to blog was not strong enough for me to
  3451. re-prioritize things in my day so that I would do it.
  3452.  
  3453. &lt;p&gt; I had plenty to say. I continue to have plenty to say. I had plenty
  3454. of important things to do, and that also continues to be true. Despite my
  3455. other responsibilities, I am making time now, and I will continue to
  3456. make time, every so often, to say things in this blog. I am being
  3457. &lt;em&gt;intentional&lt;/em&gt; about it.
  3458.  
  3459. &lt;p&gt; To be intentional about something means to be deliberately purposeful:
  3460. to make one&apos;s actions a directly chosen consequence of one&apos;s thoughtful
  3461. decisions. For most people, myself included, life is full of input,
  3462. distractions, demands, requests. It is easy to fill time without much
  3463. effort. But if I am not intentional, it will be filled with reaction,
  3464. not action: things that circumstances and prior commitments have chosen
  3465. for me, not things I have chosen for myself.
  3466.  
  3467. &lt;p&gt; Reaction is fine, even good and necessary. Many people, myself included,
  3468. build up throughout their lives various important responsibilities:
  3469. responsibilities to family, work, friends, communities. Responsibilities
  3470. carry with them a commitment to react to the needs of others. This is
  3471. well and good. But it is not enough, at least not for me. I realize that
  3472. to be authentic, I have to consider carefully what is important to me,
  3473. decide what to do about it, and then act on it. This is intentionality. I&apos;ve
  3474. decided to be intentional about blogging. Look for more
  3475. blog entries in the coming weeks.
  3476. </description>
  3477.  </item>
  3478.  <item>
  3479.    <title>A Way to Visualize Relative Masses of Things in the Solar System </title>
  3480.    <pubDate>Tue, 12 Dec 2017 13:07:00 -0500</pubDate>
  3481.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525257Ejdd/blog.cgi/2017/12/12#astromass</link>
  3482.    <category>/misc</category>
  3483.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/astromass</guid>
  3484.    <description>
  3485. Every so often we hear things in the news about the solar system: a mission
  3486. to a planet or asteroid, talk of manned missions to mars, arguments about
  3487. whether Pluto is a planet or not.  We tend to have pretty sketchy ideas of
  3488. what most bodies in the solar system are like compared to Earth.  The fact
  3489. is that they&apos;re more wildly different in size and mass than we might think.
  3490.  
  3491. &lt;P&gt;Let&apos;s look at mass.  Imagine you decide to row across San Francisco
  3492. bay in a 12-foot aluminum rowboat.  You pack a couple of suitcases, your
  3493. 15 inch Macbook Pro (can&apos;t go without connectivity) and your ipad mini,
  3494. you get in your rowboat and start rowing.  As you row, you get hungry, so
  3495. you pull out a Snickers bar.  Now imagine that the USS Nimitz, a massive
  3496. nuclear-powered aircraft carrier, passes by.  There you are, in a rowboat
  3497. with your two suitcases, your Macbook Pro, your iPad, and your Snickers bar,
  3498. alongside a huge supercarrier.
  3499.  
  3500. &lt;P&gt; Well, the mass of the sun compared to the earth is like that aircraft
  3501. carrier compared to you and your boat.  The mass of Mars is like your
  3502. two suitcases.  The mass of the moon is like your 15 inch Macbook Pro,
  3503. and the mass of Pluto is like your iPad mini.  As for the Snickers bar,
  3504. it&apos;s like Ceres, the largest of the asteroids.
  3505.  
  3506. &lt;P&gt;Now let&apos;s suppose the massive wake of the aircraft carrier tips over
  3507. your rowboat and leaves you in the water.  Along comes a rich tech founder
  3508. in his 70 foot yacht, and fishes you out.  That yacht is like Jupiter,
  3509. the largest planet.
  3510.  
  3511. &lt;P&gt;So forget any mental images you might have of planets being something
  3512. like the Sun, only a bit smaller and cooler.  The sizes of things in the
  3513. solar system are really quite different, and there is nothing, absolutely
  3514. nothing, in the solar system that is anything quite like the Sun.
  3515. </description>
  3516.  </item>
  3517.  <item>
  3518.    <title>Bitcoin, Cryptocurrency and Blockchain</title>
  3519.    <pubDate>Mon, 11 Dec 2017 14:02:00 -0500</pubDate>
  3520.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525257Ejdd/blog.cgi/2017/12/11#bitcoin</link>
  3521.    <category>/it</category>
  3522.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/bitcoin</guid>
  3523.    <description>
  3524. &lt;P&gt;As the price of Bitcoin goes up and up, talk increases about Bitcoin
  3525. and other cryptocurrencies, like Litecoin, Monero, ZCash, Ethereum,
  3526. and many others. Plenty is being said, and it can be a bit confusing.
  3527.  
  3528. &lt;P&gt;But there is no need to be confused.  Bitcoin and other cryptocurrencies
  3529. are basically simple.  They are not coins.  They are simply lists.
  3530. Each cryptocurrency has a master list.  The list typically contains
  3531. information about &lt;em&gt;who&lt;/em&gt; and &lt;em&gt;what&lt;/em&gt; (i.e. amounts).  The list
  3532. is designed in a clever way, using computer software, so that people
  3533. all over the world can have identical copies of the list and keep it up
  3534. to date, without someone having to be the holder of the &quot;master copy&quot;.
  3535. But it is still just a list.  
  3536.  
  3537. &lt;P&gt; The sort of list used for cryptocurrencies is called a &quot;blockchain&quot;,
  3538. and it has some special properties.  One particularly clever property is
  3539. that you can&apos;t normally just add anything you want to the list, there is a
  3540. scheme to control that.  Instead, you need to arrange with someone already
  3541. on the list to give up (some of) their place on the list to you.
  3542.  
  3543. &lt;P&gt;So when someone says they bought some Bitcoin and they&apos;re going to make
  3544. a lot of money, what they mean (whether they realize it or not) is that
  3545. they paid somebody some money to put them on a list, and they hope that
  3546. someone later will pay them even more money to get off it.  
  3547.  
  3548. &lt;P&gt;As for me, I haven&apos;t &quot;bought&quot; any. As I write this, cryptocurrency
  3549. prices are rising fast.  But I think what is happening is a kind of run-away
  3550. positive feedback loop: people are buying in because it is going up, and it
  3551. is going up because people are buying in.  Eventually it will run out of
  3552. people to buy in, and it will stop going up.  Then some people will sell,
  3553. causing the feedback loop to go the other way: people will sell because
  3554. it is going down, and it will go down because people are selling.
  3555.  
  3556. &lt;P&gt;That being said, one thing in particular about cryptocurrency is
  3557. making me grumpy about it, even though I don&apos;t &quot;own&quot; any. Recall I wrote
  3558. that you can&apos;t &lt;em&gt;normally&lt;/em&gt; make yourself a new entry on a blockchain
  3559. list, but there is a way.  You can do an enormous lot of computations on
  3560. a computer in an attempt to find new special numbers that can be used
  3561. to create new entries on the list. This process is misnamed &quot;mining&quot;,
  3562. but it&apos;s more a sort of computerized brute-force mathematical searching.
  3563. Those computations take a long time and use a lot of electricity.  Moreover,
  3564. even the ordinary transactions generated by people &quot;buying&quot; and &quot;selling&quot;
  3565. a cryptocurrency is a computational burden, since there
  3566. are so many copies of the list around the world.  Each list
  3567. is very big: Bitcoin&apos;s is more than 100GB, and every copy
  3568. need to be updated.  This uses electricity too. In fact, &lt;A
  3569. HREF=&quot;https://digiconomist.net/bitcoin-energy-consumption&quot;&gt;digiconomist.net
  3570. estimates that Bitcoin computations alone presently use up enough
  3571. electricity to power more than three million US households.&lt;/A&gt;
  3572. Furthermore, the &quot;mining&quot; computers use GPUs that are
  3573. really good for graphics and machine learning, but because cryptocurrency
  3574. &quot;miners&quot; are buying them all up, those GPUs are getting harder to
  3575. find for a good price. Personally, I am not happy with the
  3576. challenges I am having in finding enough GPU resources for our computer
  3577. scientists, who are hungry for GPUs for machine learning.  While high
  3578. demand for GPUs is maybe good for GPU manufacturers (for example, &lt;A
  3579. HREF=&quot;http://fortune.com/2017/09/03/bitcoin-graphics-card-makers-stocks/&quot;&gt;according
  3580. to fortune.com, Nvidia made U$150M in one quarter in 2017 selling GPUs
  3581. to cryptocurrency &quot;miners&quot;&lt;/A&gt;), surely all those GPUs, and all that
  3582. electricity, can be used for something more useful than cryptocurrency.
  3583. </description>
  3584.  </item>
  3585.  <item>
  3586.    <title>A closer look at topuniversities.com&apos;s 2017 rankings for Computer Science.</title>
  3587.    <pubDate>Thu, 09 Mar 2017 12:58:00 -0500</pubDate>
  3588.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2017/03/09#rankings2017</link>
  3589.    <category>/misc</category>
  3590.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/rankings2017</guid>
  3591.    <description>
  3592. &lt;p&gt;&lt;A HREF=&quot;http://topuniversities.com&quot;&gt;The QS World University Rankings&lt;/A&gt;
  3593. for 2017 are out, including the subject rankings.  For the subject &quot;Computer
  3594. Science &amp; Information Systems&quot;, the University of Toronto does very well,
  3595. placing tenth.
  3596.  
  3597. &lt;P&gt;A closer look at the top ten shows some expected leaders (MIT, Stanford,
  3598. CMU, UC Berkeley) but some less expected ones, such as Oxford and Cambridge.
  3599. These are superb Universities with good Computer Science programs, but
  3600. are their CS programs really among the ten best in the world?
  3601.  
  3602. &lt;P&gt;A closer look at how the score is computed sheds some light on this
  3603. question.  The Overall Score is a combination of Academic Reputation,
  3604. Citations per Paper, Employer Reputation, and H-index Citations.
  3605. Academic Reputation and Employer Reputation are, in essence, the opinions of
  3606. professors and employers respectively. While (hopefully) they are reasonably
  3607. well founded opinions, this is a subjective, not an objective, metric.
  3608. On the other hand, Citations per Paper and H-index Citations are objective.
  3609. So I looked at Citations per Paper and H-index Citations for the top forty
  3610. schools on the 2017 QS Computer Science &amp; Information Systems ranking.
  3611.  
  3612. &lt;P&gt;By Citations per Paper, top five of those forty are:
  3613. &lt;OL&gt;
  3614. &lt;LI&gt;Princeton
  3615. &lt;LI&gt;Stanford
  3616. &lt;LI&gt;UT Austin
  3617. &lt;LI&gt;Washington
  3618. &lt;LI&gt;UC Berkeley
  3619. &lt;/OL&gt;
  3620.  
  3621. &lt;P&gt;No MIT?  This seems off.  So lets look at the top five by H-Index Citations:
  3622.  
  3623. &lt;OL&gt;
  3624. &lt;LI&gt;Stanford
  3625. &lt;LI&gt;MIT
  3626. &lt;LI&gt;UC Berkeley
  3627. &lt;LI&gt;UI Urbana-Champaign
  3628. &lt;LI&gt;UT Austin
  3629. &lt;/OL&gt;
  3630.  
  3631. &lt;P&gt;That looks more reasonable.  So let&apos;s look at the top twenty by H-Index Citations:
  3632. &lt;OL&gt;
  3633. &lt;LI&gt;Stanford
  3634. &lt;LI&gt;MIT
  3635. &lt;LI&gt;UC Berkeley
  3636. &lt;LI&gt;UI Urbana-Champaign
  3637. &lt;LI&gt;UT Austin
  3638. &lt;LI&gt;Georgia IT
  3639. &lt;LI&gt;CMU
  3640. &lt;LI&gt;Tsinghua
  3641. &lt;LI&gt;Nanyang
  3642. &lt;LI&gt;ETH Zurich
  3643. &lt;LI&gt;Washington
  3644. &lt;LI&gt;Princeton
  3645. &lt;LI&gt;UBC
  3646. &lt;LI&gt;Toronto
  3647. &lt;LI&gt;Waterloo
  3648. &lt;LI&gt;NU Singapore
  3649. &lt;LI&gt;UC London
  3650. &lt;LI&gt;Cornell
  3651. &lt;LI&gt;UCLA
  3652. &lt;LI&gt;CU Hong Kong
  3653. &lt;/OL&gt;
  3654.  
  3655. &lt;P&gt;That&apos;s a list that makes more sense to me.  While it puts my department
  3656. 14th instead of 10th, I think I have more confidence in the objectivity
  3657. of this ordering than I do in the QS Overall Score ordering.
  3658. </description>
  3659.  </item>
  3660.  <item>
  3661.    <title>Program Source Code Should be Readable by Human Beings By Definition</title>
  3662.    <pubDate>Thu, 02 Feb 2017 13:35:00 -0500</pubDate>
  3663.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2017/02/02#python3</link>
  3664.    <category>/it</category>
  3665.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/python3</guid>
  3666.    <description>
  3667. Version 3 of the &lt;A HREF=&quot;https://www.python.org&quot;&gt;Python programming
  3668. language&lt;/A&gt; made a seemingly innocuous change to the Python programming
  3669. language: &lt;A HREF=&quot;https://www.python.org/dev/peps/pep-0008/#id18&quot;&gt;no
  3670. longer could tabs and spaces be mixed for indentation: either tabs must be
  3671. used exclusively, or spaces.&lt;/A&gt; Hence the following is not a valid Python
  3672. 3 program:
  3673. &lt;pre&gt;
  3674. def hello():
  3675. print(&quot;Hello&quot;)
  3676.        print(&quot;World&quot;)
  3677. hello()
  3678. &lt;/pre&gt;
  3679. If I run it, here&apos;s what I get:
  3680. &lt;pre&gt;
  3681. % python3 testme.py
  3682.  File &quot;testme.py&quot;, line 3
  3683.    print(&quot;World&quot;)
  3684.                 ^
  3685. TabError: inconsistent use of tabs and spaces in indentation
  3686. &lt;/pre&gt;
  3687. However, the following is a valid Python 3 program:
  3688. &lt;pre&gt;
  3689. def hello():
  3690.        print(&quot;Hello&quot;)
  3691.        print(&quot;World&quot;)
  3692. hello()
  3693. &lt;/pre&gt;
  3694. &lt;pre&gt;
  3695. % python3 testme.py
  3696. Hello
  3697. World
  3698. &lt;/pre&gt;
  3699. and so is the following:
  3700. &lt;pre&gt;
  3701. def hello():
  3702. print(&quot;Hello&quot;)
  3703. print(&quot;World&quot;)
  3704. hello()
  3705. &lt;/pre&gt;
  3706. &lt;pre&gt;
  3707. % python3 testme.py
  3708. Hello
  3709. World
  3710. &lt;/pre&gt;
  3711.  
  3712. Confused yet?  
  3713.  
  3714. &lt;p&gt;As you can, or perhaps more to the point, &lt;em&gt;can&apos;t&lt;/em&gt; see, the problem
  3715. here is that the first program uses a tab to indent the first
  3716. &lt;em&gt;print&lt;/em&gt; statement, and spaces to indent the second &lt;em&gt;print&lt;/em&gt;
  3717. statement.  The second program uses spaces to indent both, and the third
  3718. program uses tabs to indent both.  But because tabs and spaces are both
  3719. visually represented as whitespace, it is difficult or impossible to
  3720. visually distinguish between a correct and an incorrect python3 program
  3721. through inspecting the source code.  This breaks the basic definition of
  3722. source code: &lt;em&gt;human-readable computer instructions&lt;/em&gt;.
  3723.  
  3724. &lt;P&gt;No doubt the Python 3 designers have good intentions: to help
  3725. python programmers be consistent about indentation.  But to me, it
  3726. seems unreasonable to have a programming language where syntactically
  3727. or semantically important distinctions are not clearly visible in the
  3728. source code.
  3729. </description>
  3730.  </item>
  3731.  <item>
  3732.    <title>Slow Windows Update on Windows 7 again?  Install two Windows Update patches first.</title>
  3733.    <pubDate>Wed, 23 Nov 2016 09:48:00 -0500</pubDate>
  3734.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/11/23#wupdate2</link>
  3735.    <category>/it</category>
  3736.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/wupdate2</guid>
  3737.    <description>
  3738. Back in May, &lt;A
  3739. HREF=&quot;http://www.cs.toronto.edu/~jdd/blog.cgi/2016/05/09#wupdate&quot;&gt;
  3740. I wrote&lt;/A&gt; about Windows Update for Windows 7 taking many hours or even
  3741. days; the fix then was to install two patches manually first.
  3742.  
  3743. &lt;P&gt; The problem has returned. Even if you install the two patches I mentioned
  3744. in May, you may experience very slow updates on Windows 7.
  3745.  
  3746. &lt;P&gt;Happily, again there&apos;s a workaround: grab two patches,
  3747. different than before, and manually install them. Get &lt;A
  3748. HREF=&quot;https://support.microsoft.com/en-ca/kb/3172605&quot;&gt;
  3749. KB3172605&lt;/A&gt; and its prerequisite &lt;A
  3750. HREF=&quot;https://support.microsoft.com/en-ca/kb/3020369&quot;&gt;KB3020369&lt;/A&gt; from the
  3751. &lt;A HREF=&quot;https://www.microsoft.com/download&quot;&gt;Microsoft Download Center&lt;/A&gt;,
  3752. and install them manually in numeric order, before running Windows update.
  3753. If making a fresh Windows 7 installation, simply install Windows 7 SP1,
  3754. followed by KB3020369, then KB3172605, and only then run windows update.
  3755. These two patches seem to address the slowness issues: after they were
  3756. installed on some of our systems here, Windows Update ran in a reasonable
  3757. amount of time.
  3758. </description>
  3759.  </item>
  3760.  <item>
  3761.    <title>Remembering Kelly Gotlieb</title>
  3762.    <pubDate>Wed, 26 Oct 2016 10:41:00 -0400</pubDate>
  3763.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/10/26#gotlieb</link>
  3764.    <category>/misc</category>
  3765.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/gotlieb</guid>
  3766.    <description>
  3767. &lt;P&gt;On October 16th, 2016, &lt;A
  3768. HREF=&quot;http://web.cs.toronto.edu/news/current/In_Memoriam__The__father_of_computing_in_Canada__Calvin_C__Gotlieb.htm&quot;&gt;Kelly
  3769. Gotlieb&lt;/A&gt;, founder of the &lt;A HREF=&quot;http://www.cs.toronto.edu&quot;&gt;Department
  3770. of Computer Science&lt;/A&gt; at the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University
  3771. of Toronto&lt;/A&gt;, passed away in his 96th year.  I had the privilege
  3772. of knowing him.  Kelly was a terrific person: brilliant, kind, and
  3773. humble. He was always willing to make time for people.  He was a great
  3774. thinker: his insights, particularly in the area of computing and society,
  3775. were highly influential.  I never fully realized how influential
  3776. he was until we, here at the department of Computer Science,
  3777. created a &lt;A HREF=&quot;http://socialissues.cs.toronto.edu&quot;&gt;blog,
  3778. http://socialissues.cs.toronto.edu&lt;/A&gt;,
  3779. in honour of the 40th anniversary of &lt;em&gt;
  3780. &lt;A HREF=&quot;https://books.google.ca/books?id=xLTiBQAAQBAJ&quot;&gt;Social
  3781. Issues in Computing&lt;/A&gt;&lt;/em&gt;, the seminal textbook he and &lt;A
  3782. HREF=&quot;http://www.cs.toronto.edu/~bor&quot;&gt;Allan Borodin&lt;/A&gt; wrote in 1973
  3783. in the area of computers and society.  I served as editor of the blog,
  3784. and solicited contributions from the top thinkers in the field. So many
  3785. of them responded, explaining to me how influential his ideas had been to
  3786. them, and the blog was filled with insightful articles building in various
  3787. ways upon the foundation that he and Allan had laid so many years before. I
  3788. interviewed Kelly for the blog, and he was terrific: even in his nineties,
  3789. he was full of insights. His mind active and enthusiastic, he was making
  3790. cogent observations on the latest technologies, ranging from self-driving
  3791. cars to automated medical diagnosis and treatment.
  3792.  
  3793. &lt;P&gt;To me, Kelly epitomized the truth about effective teaching that is all too
  3794. often missed: teaching is not just about information, teaching is about
  3795. inspiration.  Kelly was a truly inspiring teacher and thinker.  He was
  3796. completely authentic in everything he did, he was full of enthusiasm,
  3797. and that enthusiasm was infectious.  Conversations with Kelly so often
  3798. left me energized and inspired, thinking along new directions of thought
  3799. that something he said had triggered, or leaping past obstacles that
  3800. had previously seems insurmountable.  That is true teaching. Information
  3801. without inspiration is simply fodder for forgetfulness, but teaching that
  3802. inspires leads to new insights, integration of ideas, genuine understanding,
  3803. and a better, clearer and sharper window on the world. Kelly inspired so
  3804. many people for so many years.  We are truly blessed that he was among us.
  3805. He will be remembered.
  3806. </description>
  3807.  </item>
  3808.  <item>
  3809.    <title>The Price of Google</title>
  3810.    <pubDate>Sun, 16 Oct 2016 18:02:00 -0400</pubDate>
  3811.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/10/16#google</link>
  3812.    <category>/it</category>
  3813.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/google</guid>
  3814.    <description>
  3815. I am a Canadian still living in the city in which I was
  3816. born. I love living in Canada, but life in Canada has its price.
  3817. &lt;A HREF=&quot;http://thecanadianencyclopedia.ca/en/article/al-purdy/&quot;&gt;Al
  3818. Purdy&lt;/A&gt;, the late 20th century Canadian poet, once wrote about Canada
  3819. as a country where everyone knows, but nobody talks about, the fact that
  3820. you can die from simply being outside.  It is true, of course: almost
  3821. everywhere in Canada, the winter is cold enough that a sufficient number
  3822. of hours outside without protection can lead to death by exposure. But
  3823. this basic fact is designed into pretty much everything in Canadian life,
  3824. it is simply accepted as a given by well over thirty million Canadians,
  3825. and we cope: we wear the right winter clothes, we heat and insulate our
  3826. buildings in winter, we equip our cars with the right tires, and life goes
  3827. on. Despite the Canadian winter, Canada is a great place to live.
  3828.  
  3829. &lt;P&gt;Google offers a lot of very good free web services: it is &quot;a great
  3830. place to live&quot; on the Internet, and their services are used by hundreds
  3831. of milliions of people all over the world.  While Google seems about as
  3832. far removed from a Canadian winter as you can imagine, there&apos;s something
  3833. in their Terms of Service that people seem to rarely talk about, something
  3834. that might have a bit of a chilling effect on one&apos;s initial ardor.
  3835.  
  3836. &lt;P&gt;Google, to its credit, has a very clear and easy-to-read &lt;A
  3837. HREF=&quot;https://www.google.com/intl/en/policies/terms/&quot;&gt;Terms of
  3838. Service document&lt;/A&gt;. Here&apos;s an excerpt from the version of April 14, 2014,
  3839. which is the most current version at the time I write this.
  3840.  
  3841. &lt;blockquote&gt;
  3842. When you upload, submit, store, send or receive content to or through our
  3843. Services, you give Google (and those we work with) a worldwide license to
  3844. use, host, store, reproduce, modify, create derivative works (such as those
  3845. resulting from translations, adaptations or other changes we make so that
  3846. your content works better with our Services), communicate, publish, publicly
  3847. perform, publicly display and distribute such content. The rights you grant
  3848. in this license are for the limited purpose of operating, promoting, and
  3849. improving our Services, and to develop new ones. This license continues
  3850. even if you stop using our Services (for example, for a business listing
  3851. you have added to Google Maps).
  3852. &lt;/blockquote&gt;
  3853.  
  3854. Let me pull out for closer examination the most important bits.
  3855. For readability, I&apos;ve omitted elipses.
  3856.  
  3857. &lt;blockquote&gt;
  3858. When you submit content to our Services, you give Google (and those we work
  3859. with) a worldwide license to use such content for the purpose of our Services.  
  3860. This continues even if you stop using our Services.
  3861. &lt;/blockquote&gt;
  3862.  
  3863. &lt;P&gt;As you can see, this is pretty broad.  You are granting Google and their
  3864. partners the right to use your content for Google&apos;s Services (present and
  3865. future) anywhere in the world, forever.  While it does say that it must be
  3866. used for the purpose of their Services, it doesn&apos;t limit itself to
  3867. existing Services and it doesn&apos;t constrain what a &quot;Service&quot; might be.  Since
  3868. developing and offering Services, broadly understood, pretty much covers
  3869. the gamut of what Google does as a company, the answer is &lt;em&gt;Yes&lt;/em&gt;:
  3870. by submitting content to their services, you are granting Google and their
  3871. partners the right to use your content anywhere in the world, forever,
  3872. for a broadly unconstrained set of purposes.
  3873.  
  3874. &lt;P&gt;So does this mean nobody should use Google?  Does the Canadian winter
  3875. mean that nobody should live in Canada? After all, as Al Purdy writes,
  3876. in Canada you can die from simply being outside.
  3877.  
  3878. &lt;P&gt;Well, no, of course not.  While Google has the right to do broadly
  3879. unconstrained things with our content that we submit to them, their self
  3880. -interest is typically aligned with our&apos;s: they want us to entrust our
  3881. content to them, because they use it to earn money to operate. Therefore,
  3882. to persuade us to keep submitting content to them, they will work hard
  3883. to protect and secure the content they already have, in ways they think
  3884. we consider important.  For this reason, I think it&apos;s not unreasonable to
  3885. trust Google with some of my content: I believe they are likely to protect
  3886. it in sensible ways.  Other content I choose not to submit to Google.
  3887. Just as I am prepared for a Canadian winter, knowing it is the price
  3888. I pay to live in Canada, I continue to use some Google services, knowing that
  3889. they will keep and use my content.  Many Google services are very good and
  3890. well worth using, much of my content is not very sensitive, and I trust
  3891. Google enough to share content with them.
  3892.  
  3893. &lt;p&gt;I do wonder, however, how many Google users really understand the rights
  3894. they are granting to Google.  Canada has been around for centuries: the
  3895. Canadian winter is no secret.  But the implications of Google&apos;s broad
  3896. right to use our content are not quite so obvious.  It&apos;s not really so clear
  3897. how Google is using the content or might use it in the future, and even if
  3898. we trust Google, can we trust all those who might put pressure on Google?
  3899. Quite frankly, we really don&apos;t know yet how Google&apos;s massive repository of
  3900. our collective content can be used.  We can envision wonderful outcomes:
  3901. historians a century or two hence coming to insightful conclusions about
  3902. early twenty-first century society, for example, but we can also envision
  3903. outcomes not quite so sanguine: for example, a twenty-first century
  3904. version of Orwell&apos;s &lt;em&gt;1984&lt;/em&gt;, a dystopian world of &quot;thought-crimes&quot;
  3905. and &quot;doublespeak&quot; where content is is scanned for dissent from a prevailing
  3906. ideology.  A certain degree of caution may be warranted: in the case of
  3907. Google, unlike Canada, we may not have yet seen how severe winter can be.
  3908. A certain degree of caution is warranted.  Yes, use Google, but use it
  3909. knowing what you are doing.
  3910.  
  3911. &lt;P&gt;One last thing to be said: I focus on Google here, but
  3912. the same issues hold for Facebook, Twitter, Yahoo and other purveyors
  3913. of free services over the Internet.  Read their Terms of Service to
  3914. learn what rights you are granting by your use of their services,
  3915. and decide on the basis of that knowledge how to use their services,
  3916. and even whether you use their services at all.  After all, &lt;A
  3917. HREF=&quot;https://en.wikipedia.org/wiki/Snowbird_(person)&quot;&gt;even Canadians
  3918. sometimes choose to spend winter in Florida, Mexico, or Arizona&lt;/A&gt;.
  3919. </description>
  3920.  </item>
  3921.  <item>
  3922.    <title>The Sun-Managers Mailing list: a Knowledge Sharing Success Story</title>
  3923.    <pubDate>Mon, 16 May 2016 20:29:00 -0400</pubDate>
  3924.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/05/16#sunmanagers</link>
  3925.    <category>/it</category>
  3926.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/sunmanagers</guid>
  3927.    <description>
  3928. &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu&quot;&gt;Sun-Managers&lt;/A&gt; was an email
  3929. mailing list for system administrators of computers made by &lt;A
  3930. HREF=&quot;http://en.wikipedia.org/wiki/Sun_Microsystems&quot;&gt;Sun Microsystems,
  3931. Inc&lt;/A&gt;.  The list operated from mid-1989 to the fall of 2014, and I was
  3932. privileged to be part of it for almost all of its history.
  3933.  
  3934. Sun-Managers was founded in May of 1989 by &lt;A
  3935. HREF=&quot;https://www.linkedin.com/in/william-lefebvre-936b76&quot;&gt;William (Bill)
  3936. LeFebvre&lt;/A&gt;, at &lt;A HREF=&quot;http://www.northwestern.edu&quot;&gt;Northwestern
  3937. University&lt;/A&gt;.  At the time, Bill ran &lt;A
  3938. HREF=&quot;http://ftp.lanet.lv/ftp/sun-info/sunflash/1990/Aug/20.10.sunspots&quot;&gt;Sun-Spots&lt;/A&gt;,
  3939. a digest-format mailing list for system administrators
  3940. of Sun systems, but the digest format made it difficult
  3941. for people to ask questions and get a timely response. He &lt;A
  3942. HREF=&quot;https://groups.google.com/forum/#!msg/comp.sys.sun/H9z8YwunvB4/xJFp8qOLTFUJ&quot;&gt;created
  3943. Sun-Managers&lt;/A&gt;, an unmoderated mailing list intended for
  3944. short-turnaround time questions.  This was an immediate success:
  3945. so much so that by the fall of 1989, the sheer number of messages on
  3946. the list were swamping mailboxes.  In Nov 1989, Bill instituted a &lt;A
  3947. HREF=&quot;http://sunmanagers.cs.toronto.edu/policy.txt&quot;&gt;simple policy&lt;/A&gt;:
  3948. if someone asks a question on the list, other list members were expected
  3949. to reply by email directly to the person asking the question, not to the
  3950. list. The person asking the question, in turn, was expected to summarize
  3951. the answers received, and send the summary to the list.
  3952.  
  3953. &lt;P&gt; I joined the list about this time: I had started a new job at
  3954. the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of Toronto&lt;/A&gt;&apos;s &lt;A
  3955. HREF=&quot;http://www.cs.toronto.edu&quot;&gt;Computer Science department&lt;/A&gt;, a role that
  3956. included the administration of a number of Sun workstations and servers.
  3957. I was looking for resources to help me with my Sun system administration
  3958. tasks, and this list was an excellent one.  Because of this summary policy, the
  3959. list volume was manageable enough that I could keep up, yet the turnaround
  3960. time on questions was short.  I mostly &quot;lurked&quot; at first, reading but not
  3961. replying. I felt too inexpert to answer many questions, and too shy to ask.
  3962. However, I learned a great deal from what I read. Moreover, the summaries
  3963. were archived, and &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/year.html&quot;&gt;
  3964. this archive&lt;/A&gt; became a resource in itself, a knowledge-base of practical
  3965. information about administering Sun systems.
  3966.  
  3967. &lt;p&gt; The list grew very rapidly: 343 summaries in 1990, and over 1000 in
  3968. 1991. In August of that year, it was noted that certain questions were being
  3969. asked often, and rather than waste effort answering the same question several
  3970. times, a &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/faq.html&quot;&gt;&quot;Frequently
  3971. Asked Questions&quot; (FAQ) file&lt;/A&gt; was instituted. The &lt;A
  3972. HREF=&quot;http://sunmanagers.cs.toronto.edu/1991/0484.html&quot;&gt;first version
  3973. &lt;/A&gt; was created by a list member from &lt;A HREF=&quot;http://www.bu.edu&quot;&gt;Boston
  3974. University&lt;/A&gt;, and quickly grow to dozens of answers.
  3975.  
  3976. &lt;p&gt; By November of 1992, the list had grown to thousands of members,
  3977. and the workload of managing the list, editing the FAQ and coaching
  3978. list members on how to follow the &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/policy.txt&quot;&gt;list policy&lt;/A&gt;
  3979. had become significant. Many list members were not individuals,
  3980. but &quot;mail exploders&quot;: email addresses that themselves were mailing
  3981. lists going to multiple individuals at a given site.  This made
  3982. handling list membership issues more complex.  Bill LeFebvre decided
  3983. to hand the list over to others. Two list members stepped up: &lt;A
  3984. HREF=&quot;http://www.linkedin.com/in/gene-rackow-87b7b25&quot;&gt;Gene Rackow&lt;/A&gt; from
  3985. &lt;A HREF=&quot;http://www.anl.gov&quot;&gt;Argonne National Laboratory&lt;/A&gt; to run the
  3986. list software, and me, to handle the FAQ and policy work. By this time,
  3987. I had benefitted from the list for a while, and I felt it was time to
  3988. &quot;give back&quot;.  At the time, I wasn&apos;t in a position to actually
  3989. run the list: I&apos;d just taken on a new role as system manager of
  3990. the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of Toronto &lt;/A&gt;&lt;A
  3991. HREF=&quot;http://www.cs.toronto.edu&quot;&gt;Computer Science Department&lt;/A&gt;&apos;s &lt;A
  3992. HREF=&quot;http://www.cdf.toronto.edu&quot;&gt;teaching laboratories&lt;/A&gt;, and had my
  3993. hands full, but I could certainly help with content.  I was really glad to
  3994. work together with Gene, a seasoned system administrator, on this rapidly
  3995. growing list, which we moved to a system at &lt;A
  3996. HREF=&quot;http://www.anl.gov&quot;&gt;Argonne National Labs&lt;/A&gt;, where Gene worked.
  3997.  
  3998. &lt;p&gt; The list continued to grow through the 1990s.  During this time,
  3999. Sun Microsystems was quietly supportive, helping Gene with hardware (a
  4000. &lt;A HREF=&quot;http://en.wikipedia.org/wiki/SPARCstation_1&quot;&gt;Sparcstation 1&lt;/A&gt;)
  4001. as the list grew.  By 1996, over two thousand summaries a year were being
  4002. produced, peaking at 2243 in 2002.
  4003.  
  4004. In May of 1998, Gene Rackow handed over list management to Rob
  4005. Montjoy from the University of Cincinnati, who in turn handed over
  4006. list management to &lt;A HREF=&quot;https://www.linkedin.com/in/mrbill&quot;&gt;Bill
  4007. Bradford&lt;/A&gt; in November of 2000. The list was moved from Argonne
  4008. National Labs to a system in Austin run by Bill.  I continued to manage
  4009. the &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/policy.txt&quot;&gt;list
  4010. policy&lt;/A&gt; and edit list information files, such as &lt;A
  4011. HREF=&quot;http://sunmanagers.cs.toronto.edu/before-posting.txt&quot;&gt;a
  4012. &quot;think before posting&quot; reminder&lt;/A&gt; and the FAQ which had grown to
  4013. 79 questions by December 2000. This had become a bit too large, and
  4014. so &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/faq-2000.txt&quot;&gt;19
  4015. questions deemed less frequently asked&lt;/A&gt; were
  4016. trimmed.  A further trim was made in 2005, reducing a &lt;A
  4017. HREF=&quot;http://sunmanagers.cs.toronto.edu/faq-2005.txt&quot;&gt;65-question FAQ&lt;/A&gt;
  4018. to &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/faq.txt&quot;&gt;one under 60&lt;/A&gt;.
  4019.  
  4020. &lt;p&gt; By 2002, the list had reached over five thousand members and the
  4021. workload of running the list software and managing the list
  4022. subscriptions had become too much for one person.  Dan Astoorian, my
  4023. colleage at the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of Toronto&lt;/A&gt;,
  4024. stepped in to help, and he was sorely needed.  Moreover, the list server
  4025. hardware was feeling the strain: by mid-2001, list members were being
  4026. asked to contribute used equipment to upgrade the server.  This was
  4027. resolved in April 2003, when the list was migrated to a machine at the
  4028. &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of Toronto&lt;/A&gt; that had been
  4029. donated to the University by Sun Microsystems.
  4030.  
  4031. &lt;p&gt; But times were changing.  Linux was growing rapidly and Sun&apos;s business
  4032. was being affected. The web provided more resources for people seeking
  4033. help administering their systems, and fewer were relying on mailing lists.
  4034. The list fell below 2000 summaries per year in 2003, under 1200 in 2004, and
  4035. dropped below 1000 in 2005. By 2008, summaries per year had fallen to about
  4036. 300, fewer than in any full-year period previously. Sun Microsystems ran into
  4037. significant difficulties during the economic downturn that year,
  4038. and was sold to
  4039. &lt;A HREF=&quot;http://www.oracle.com&quot;&gt; Oracle&lt;/A&gt; the following year.  As for the list, in 2009,
  4040. there were just over 200 summaries, declining to less than 100 in 2011.
  4041. More disturbingly, the ratio of summaries to questions was steadily declining,
  4042. from over 24% in 2001 to less than 16% by 2010: for some reason, list
  4043. members were becoming less diligent in summarizing responses back to the
  4044. list.  Summaries and list traffic in general continued to decline rapidly:
  4045. there were just over 50 summaries in 2012, and less than a dozen in 2013.
  4046. In 2014, there were only three by October, when a hardware failure provided
  4047. a good excuse to retire the list.
  4048.  
  4049. &lt;p&gt; The Sun-Managers mailing list, over its twenty-five year lifetime,
  4050. provided help to many thousands of system administrators, producing over
  4051. 29000 summaries, &lt;A HREF=&quot;http://sunmanagers.cs.toronto.edu/year.html&quot;&gt;an
  4052. archive of which continues to be available&lt;/A&gt;.  Special thanks is due to
  4053. the superb people I was privileged to work together with on the list over
  4054. the years: William LeFebvre, Gene Rackow, Rob Montjoy, Bill Bradford, and
  4055. Dan Astoorian.  Gratitude, also, is due to the thousands of list members
  4056. who so freely shared their knowledge and expertise with others.
  4057.  
  4058. &lt;P&gt; The list summary archive, and an account of the list&apos;s
  4059. history (on which this blog entry is based) is available at &lt;A
  4060. HREF=&quot;http://sunmanagers.cs.toronto.edu&quot;&gt;http://sunmanagers.cs.toronto.edu&lt;/A&gt;.
  4061. The list&apos;s official web page, &lt;A
  4062. HREF=&quot;http://www.sunmanagers.org&quot;&gt;http://www.sunmanagers.org&lt;/A&gt;, continues to be
  4063. maintained by Bill Bradford.
  4064. </description>
  4065.  </item>
  4066.  <item>
  4067.    <title>Slow Windows Update on Windows 7?  Install two Windows Update patches first.</title>
  4068.    <pubDate>Mon, 09 May 2016 10:54:00 -0400</pubDate>
  4069.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/05/09#wupdate</link>
  4070.    <category>/it</category>
  4071.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/wupdate</guid>
  4072.    <description>
  4073. Recently, I noticed Windows Update taking many hours or even days on
  4074. Windows 7, especially for new installs/reinstalls.  Task manager shows
  4075. svchost.exe exhibiting large memory usage (suggestive of a memory leak)
  4076. and/or sustained 100% CPU.
  4077.  
  4078. &lt;P&gt;Happily, there&apos;s a workaround: grab a couple of patches
  4079. to Windows Update itself, and manually install them. Get &lt;A
  4080. HREF=&quot;https://support.microsoft.com/kb/3050265&quot;&gt;KB3050265&lt;A&gt; and &lt;A
  4081. HREF=&quot;https://support.microsoft.com/kb/3102810&quot;&gt;KB3102810&lt;/A&gt; from
  4082. the &lt;A HREF=&quot;https://www.microsoft.com/download&quot;&gt;Microsoft Download Center&lt;/A&gt;,
  4083. and install them manually in that order, before running Windows update.
  4084. These two patches seem to address the issues: after they were installed
  4085. on some of our systems here, Windows Update ran in a reasonable amount of
  4086. time (an hour or two perhaps on slow systems when many updates are needed,
  4087. but not days).
  4088. </description>
  4089.  </item>
  4090.  <item>
  4091.    <title>Apple vs FBI: it is about setting a precedent.</title>
  4092.    <pubDate>Fri, 04 Mar 2016 10:25:00 -0500</pubDate>
  4093.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2016/03/04#applefbi</link>
  4094.    <category>/it</category>
  4095.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/applefbi</guid>
  4096.    <description>
  4097. There seems to be lots of confusion about Apple&apos;s current dispute with
  4098. the FBI, despite &lt;A HREF=&quot;http://www.apple.com/customer-letter&quot;&gt;Apple&apos;s
  4099. message to their customers of Feb 16, 2016&lt;/A&gt;, where they tried to
  4100. explain the issue.  Here&apos;s the issue in a nutshell.
  4101.  
  4102. &lt;p&gt; The FBI has an Apple iPhone that was the work-phone of a now-dead
  4103. terrorist.  The FBI wants to read what is on that phone.  But the
  4104. phone is encrypted, and runs a secure version of iOS.  The FBI wants Apple
  4105. to make an insecure version of iOS to run on that phone, so that the FBI
  4106. can break into the phone and read the contents.  Apple has, so far, refused.
  4107.  
  4108. &lt;p&gt; This issue will no doubt be addressed in the US courts and legislatures.
  4109. What is at stake is the precedent it sets.  The essential question is this:
  4110. to what extent should law enforcement be able to compel others to assist
  4111. them with an investigation?  Should software developers be expected to
  4112. make insecure versions of their software, so that law enforcement can
  4113. &quot;break in&quot;?  It will be very interesting to see how this plays out.
  4114. </description>
  4115.  </item>
  4116.  <item>
  4117.    <title>Apple&apos;s new Macbook laptop: like a tablet?</title>
  4118.    <pubDate>Fri, 13 Mar 2015 11:08:00 -0400</pubDate>
  4119.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2015/03/13#macbook</link>
  4120.    <category>/it</category>
  4121.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/macbook</guid>
  4122.    <description>
  4123. &lt;P&gt;I rarely write about Apple&apos;s products because they have no shortage of
  4124. press already: Apple has superb marketing, and many of their products are
  4125. remarkable in one way or another, often for excellent design and engineering.
  4126. Their new super-thin &lt;A HREF=&quot;http://www.apple.com/macbook/&quot;&gt;Macbook
  4127. laptop&lt;/A&gt; is no exception: it&apos;s very thin and light, has a superb
  4128. high-resolution screen, a carefully redesigned trackpad and keyboard,
  4129. and is very power-efficient.  New to this machine is the fact that it
  4130. has only a single USB-C port for power, data, and video (it also has a
  4131. headphone port for audio).  Most laptops have many more ports than this.
  4132. A USB port used for both power and data, and a headphone port, but nothing
  4133. else, is more typical of a tablet, not a laptop.
  4134.  
  4135. Indeed, some of the press seems to have
  4136. really latched onto this &quot;tablet&quot; comparison.  &lt;A
  4137. HREF=&quot;http://www.foxnews.com/tech/2015/03/11/apple-has-designed-macbook-thats-almost-tablet/&quot;&gt;Brooke
  4138. Crothers of Foxnews/Tech claims that the MacBook is &quot;almost a tablet&quot;
  4139. and states that the MacBook &quot;is an iPad with a keyboard&quot; &lt;/A&gt;while  &lt;A
  4140. HREF=&quot;http://www.slate.com/blogs/future_tense/2015/03/09/apple_releases_2_lb_12_retina_macbook.html&quot;&lt;/A&gt;Lily
  4141. Hay Newman of Slate claims that &quot;you should think of the new macbook as
  4142. a tablet&quot;&lt;/A&gt;.  So how true is this?  Is the new MacBook like a tablet?
  4143.  
  4144. &lt;P&gt;Well, no, it&apos;s not.  The MacBook&apos;s screen is not
  4145. touch-capable, and is not capable of being used like a tablet screen.
  4146. The keyboard and touchpad is an integral part of the machine: it is
  4147. not optional or detachable.  It runs a desktop/laptop operating system
  4148. (MacOSX), not a tablet operating system such as iOS.  The device is not
  4149. a tablet, it is not &quot;almost a tablet&quot;, it is not even like
  4150. a tablet.  It&apos;s a small, light, power-efficient laptop.  If it
  4151. must be compared to something, perhaps it can be compared to a &lt;A
  4152. HREF=&quot;http://www.cs.toronto.edu/~jdd/blog.cgi/2009/02/09#netbooks&quot;&gt;netbook&lt;/A&gt;,
  4153. though it has a much better keyboard, touchpad and screen, and is much
  4154. more expensive.
  4155.  
  4156. &lt;P&gt;Then what about the single I/O port?  That&apos;s simply the consequence of
  4157. the new &lt;A HREF=&quot;http://www.usb.org/developers/docs/usb_31_030215.zip&quot;&gt;USB
  4158. 3.1 specification&lt;/A&gt;, which finally allows a USB connection to deliver
  4159. enough power to power a laptop, and defines the USB-C connector, which in
  4160. addition to USB data lines, provides &quot;alternate mode&quot; data lines that can
  4161. be used for display protocols like DisplayPort.  This makes it possible for
  4162. Apple to build &lt;A
  4163. HREF=&quot;http://store.apple.com/us/product/MJ1K2AM/A/usb-c-digital-av-multiport-adapter&quot;&gt;multiport
  4164. adapters for the Macbook&lt;/A&gt; that provide video (e.g. HDMI), data (USB-A)
  4165. and charging ports, making it unnecessary to provide all those ports
  4166. separately in the laptop itself.
  4167.  
  4168. &lt;P&gt;So does this make the Macbook &quot;like a tablet&quot;?  While it is true that
  4169. tablets have been using single connectors for power and data for a long
  4170. time, this doesn&apos;t make the Macbook tablet-like.  It&apos;s not the presence of
  4171. a single shared power/data connector that makes something like a tablet,
  4172. it&apos;s the interactive screen.  Yes, a horse has four legs and is often sat
  4173. upon, but a horse is not anything like a chair.
  4174.  
  4175. &lt;P&gt; So will I be getting one of the new Macbooks? Probably not: like a fine
  4176. thoroughbred, the new Macbook is lovely but rather too expensive for me.
  4177. The need to buy the multiport adapter separately makes the already high
  4178. cost of acquisition even higher.  The high price doesn&apos;t stop me from
  4179. admiring the design and engineering of this new laptop, but it does keep
  4180. me from buying one.
  4181. </description>
  4182.  </item>
  4183.  <item>
  4184.    <title>What&apos;s wrong with Blackberry? (and some ideas about how to fix it)</title>
  4185.    <pubDate>Sat, 05 Oct 2013 17:03:00 -0400</pubDate>
  4186.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2013/10/05#blackberry</link>
  4187.    <category>/it</category>
  4188.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/blackberry</guid>
  4189.    <description>Blackberry is in the news a fair bit these days, and the news seems to be all bad.  
  4190. &lt;A
  4191. HREF=&quot;http://press.blackberry.com/financial/2013/blackberry-reports-second-quarter-fiscal-2014-results.html&quot;&gt;As
  4192. the firm reports close to a billion dollars in quarterly losses&lt;/a&gt;, a &lt;A
  4193. HREF=&quot;http://www.computerworld.com/s/article/9242767/Update_Gartner_tells_IT_shops_that_it_s_game_over_for_BlackBerry&quot;&gt;Gartner
  4194. analyst recommends that enterprise customers find alternatives to Blackberry
  4195. over the next six months&lt;/A&gt;.  What&apos;s the problem?
  4196. &lt;p&gt;
  4197. Basically, fewer and fewer people want to buy Blackberry phones.  The problem
  4198. isn&apos;t so much that Blackberries don&apos;t do what they&apos;re supposed to, it&apos;s
  4199. that people now perceive iPhones and various Android phones as much better
  4200. choices, and are buying those instead.  Why?  The reason is that an iPhone or
  4201. an Android phone isn&apos;t the same sort of phone as a traditional Blackberry. An
  4202. iPhone or an Android phone is a true smartphone, i.e. an &quot;app&quot; phone, a
  4203. platform that runs a whole &quot;ecosystem&quot; of third party software.  A traditional
  4204. Blackberry is a &quot;messaging&quot; phone, a device that specializes in effective
  4205. messaging, such as email.  Yes, it can run applications too, but that&apos;s
  4206. not its primary function, and it shows.  
  4207. &lt;p&gt;
  4208. To illustrate, consider email.  Sending email requires the ability to
  4209. type quickly.  A physical keyboard works best for this, one that stretches
  4210. across the short side of the phone.  The screen, located above the keyboard,
  4211. then becomes roughly square: it can&apos;t be very wide, because the phone will
  4212. then become too wide to hold easily or to fit in one&apos;s pocket, and it can&apos;t
  4213. be very tall or the phone will become too long.   A square screen is fine
  4214. for messaging, but for other things that a smartphone might like to do,
  4215. such as displaying video, one wants a screen that is significantly wider
  4216. than it is tall.  A smartphone handles this by having a rectangular screen:
  4217. when doing messaging, one holds the phone vertical: the bottom half of the
  4218. screen then turns into a keyboard, and the top half turns into a roughly
  4219. square messaging display.  When watching media, such as videos, the phone
  4220. is held horizontal, allowing a screen that is wider than it is tall.
  4221. Hence the smartphone is useful in a broader set of ways: it is not just a
  4222. messaging device.  Smartphones have become good enough at messaging that
  4223. many people do not feel they need a dedicated messaging device.  Once the
  4224. smartphone is the only device that people feel they need to carry, there&apos;s
  4225. much less demand for a messaging phone.
  4226.  
  4227. &lt;p&gt;
  4228. Blackberry realized the problem, and tried to create a smartphone of its
  4229. own. For instance, in 2008, it released the Blackberry Storm.  But it
  4230. became clear that Blackberry&apos;s phone OS was not as well suited for
  4231. general smartphone use as iOS and Android. The Storm was not a commercial
  4232. success because it did not work as well as competing phones.  In response,
  4233. in 2010 Blackberry bought a company called QNX that had a powerful OS,
  4234. and started building devices to use it: first the Playbook, released in
  4235. spring 2011, and then the Z10 phone in early 2013, followed a few months
  4236. later by the Q10 and other phone models.
  4237.  
  4238. &lt;p&gt;
  4239. The new Blackberry OS works better than the old in delivering
  4240. smartphone apps, but it was not very mature in 2011, and was
  4241. available only on a tablet (the Blackberry Playbook). Unfortunately,
  4242. the Playbook did not sell particularly well because Blackberry &lt;A
  4243. HREF=&quot;http://press.blackberry.com/press/2010/pressrelease-4577.html&quot;&gt;
  4244. badly misrepresented it, calling it the &quot;best professional-grade table in
  4245. the industry&quot;&lt;/A&gt; though it lacked many features of the market-leading iPad,
  4246. including key messaging features such as a standalone email client.
  4247. While it could have been a market success if it were marketed as a
  4248. Blackberry phone accessory, a role it could effectively play, at release
  4249. it was clearly not a true general-purpose tablet like the iPad.  So it
  4250. accumulated few apps, while Apple&apos;s iOS and Google&apos;s Android accumulated
  4251. many.  Blackberry realized this fairly quickly, and released an Android
  4252. application emulation environment for their OS in early 2012, which allowed
  4253. many Android apps to be easily moved over to the new OS.  But few Android
  4254. developers bothered to make Blackberry versions of their Android apps,
  4255. given the relatively few Playbooks sold.
  4256.  
  4257. &lt;p&gt; In the meanwhile, Blackberry did itself no favours by making it clear
  4258. that there was no future for its existing phones, while failing to deliver
  4259. a phone running its new OS for more than a year.  This merely encouraged
  4260. Blackberry users and app developers alike to switch to another platform.
  4261. When the Z10 phone finally came out in 2013, the bulk of its apps were
  4262. those that had been written for or ported to the Playbook, a far less rich set
  4263. of applications than any Android or iOS phone.  And while the Z10 is a decent
  4264. phone that comes with some very nice messaging features, Blackberry did
  4265. not do an effective job of touting the unique features of the Z10 that
  4266. iPhones and Android phones do not have.  Moreover, the price was set high
  4267. (about the same as an iPhone or high end Android phone) and Blackberry
  4268. produced a huge number, expecting to sell a great many.  Some sold, but
  4269. many didn&apos;t, and Blackberry&apos;s recent $1B loss was due primarily to writing
  4270. down the value of unsold Z10s.
  4271.  
  4272. &lt;p&gt;
  4273. Blackberry sits today in a difficult position.  No, it is not about to go
  4274. out of business: the company is debt-free and has a couple of billion dollars
  4275. in the bank.  But its smartphone is not selling.  What should it do now?
  4276.  
  4277. &lt;p&gt;
  4278. Blackberry&apos;s best chance at this point to make its smartphone platform
  4279. viable is to take its large inventories of written-down Z10 phones and
  4280. sell them cheaply, using a renewed marketing campaign that focuses on the
  4281. unique features of the phone&apos;s software.  The Z10 hardware is really no
  4282. different than the various Android and iPhone models out there: if the
  4283. phone is to sell, it has to be on the basis of what makes it unique, and
  4284. that&apos;s the Blackberry OS software.  For instance, Blackberry should show
  4285. everyone the clever virtual keyboard that supports fast one-handed typing,
  4286. the unique messaging hub, and the &quot;Blackberry Balance&quot; software that lets
  4287. you separate work items from personal items on the phone.  Blackberry needs
  4288. to hire the best marketing people in the world to help get the message out.
  4289. This is a &quot;make or break&quot; situation for the platform.
  4290.  
  4291. &lt;p&gt;
  4292. Secondly, Blackberry should modify the OS to run Android apps natively,
  4293. without repackaging.  Android app developers are not going to repackage
  4294. their apps for Blackberry.  Blackberry needs to recognize this and make
  4295. sure that Android apps will appear automatically on Blackberry devices.
  4296. Blackberry will need to find a way to get Google Play (the Android app store)
  4297. ported to the platform.  It is too late to build a separate app ecosystem
  4298. around the Blackberry OS: it has to leverage an existing ecosystem, or die.
  4299. Android is really the only viable option for Blackberry right now.
  4300.  
  4301. &lt;p&gt;
  4302. Finally, Blackberry needs to recognize that a niche market for dedicated
  4303. messaging devices exists, and continue making devices that are the best
  4304. messaging phones available, while tapping into an existing app ecosystem.
  4305. Blackberry needs to be careful not to compromise the devices&apos; effectiveness
  4306. for messaging: it should pay attention to how people use the devices in the
  4307. real world, and address quickly whatever issues they have.  If Blackberry
  4308. can&apos;t find a way of building such messaging devices using its own OS, it
  4309. should switch to Android.  Blackberry knows how to make superb messaging
  4310. phones, and it should find a way to continue to do what it does best.
  4311. </description>
  4312.  </item>
  4313.  <item>
  4314.    <title>Cloud Computing: Everything Old is New Again</title>
  4315.    <pubDate>Tue, 20 Aug 2013 22:45:00 -0400</pubDate>
  4316.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2013/08/20#cloud</link>
  4317.    <category>/it</category>
  4318.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/cloud</guid>
  4319.    <description>
  4320. There is a great deal of hype about Cloud Computing at the moment, and
  4321. it&apos;s getting a great deal of
  4322. attention.  It&apos;s no wonder: when firms such as &lt;A
  4323. HREF=&quot;http://www.bloomberg.com/news/2013-07-21/netflix-rule-as-no-1-s-p-performer-but-.html&quot;&gt;Netflix,
  4324. with a market capitalization of over U$15B&lt;/A&gt;, use cloud computing to
  4325. deliver streaming video services to nearly forty
  4326. million customers around the world, and when the &lt;A
  4327. HREF=&quot;http://www.businessinsider.com/amazon-vs-ibm-in-a-battle-for-the-cloud-2013-7&quot;&gt;US
  4328. Central Intelligence Agency spends U$600M for cloud computing services&lt;/A&gt;,
  4329. people take notice.  But what is it all about?
  4330.  
  4331. &lt;p&gt; Cloud computing is not really a new thing, it&apos;s a variation of a very
  4332. old idea, with a new name.  In the 1960s, when computers were large and
  4333. expensive, not everyone could afford their own.  Techniques for sharing
  4334. computers were developed, and firms arose whose business was selling
  4335. time on computers to other firms.  This was most commonly described as
  4336. &quot;timesharing&quot;.  IBM released its &lt;A HREF=&quot;http://www.vm.ibm.com/history/&quot;&gt;VM
  4337. virtualization environment in 1972&lt;/A&gt;, which allowed a mainframe computer
  4338. to be divided up into virtual computers, each for a different workload.
  4339. A timesharing vendor could buy and operate an IBM computer, then rent to
  4340. their customers &quot;virtual computers&quot; that ran on that machine.  From the
  4341. customer&apos;s perspective, it was a way to obtain access to computing without
  4342. buying one&apos;s own computer.  From the vendor&apos;s perspective, it was a way of
  4343. &quot;renting out&quot; one&apos;s investment in computer infrastructure, as a viable
  4344. business.
  4345.  
  4346. &lt;p&gt;Today, cloud computing, as did timesharing in the past, involves
  4347. the renting of virtual computers to customers.  The name has changed: then,
  4348. it was called &quot;timesharing&quot;; now, &quot;cloud computing&quot;.  The type of physical
  4349. machine has changed: then, a mainframe was used to provide computing
  4350. services; now, a grid computer. The interconnection has changed: then,
  4351. leased data lines were typically used; now, the internet.  But the basic
  4352. concept is the same: a vendor rents virtual computers to customers, who
  4353. then use the virtual computers for their computing, rather than buying
  4354. their own physical computers.
  4355.  
  4356. &lt;p&gt;The advantages and disadvantages of today&apos;s cloud computing echo the
  4357. pros and cons of yesterday&apos;s timesharing.  Advantages include risk sharing,
  4358. the ability to pay for just the amount of computing needed, the option to
  4359. scale up or down quickly, the option to obtain computing resources without
  4360. having to develop and maintain expertise in operating and maintaining
  4361. those resources, and the ability to gain access to computing resources in
  4362. very large or very small quantities very quickly and easily.  Moreover,
  4363. cloud computing vendors can develop economies of scale in running physical
  4364. computers and data centres, economies that they can leverage to decrease the
  4365. cost of computing for their customers.  Disadvantages of cloud computing
  4366. include possibly higher unit costs for resources (for example, cloud
  4367. data storage and data transfer can be very expensive, especially in
  4368. large quantities), a critical dependance on the cloud computing vendor,
  4369. variable computing performance, substantial security and privacy issues,
  4370. greater legal complexity, and so on.  These tradeoffs are neither
  4371. surprising nor particularly new: in fact, many are typical of &quot;buy&quot;
  4372. vs. &quot;rent&quot; decisions in general.
  4373.  
  4374. &lt;p&gt; Then why does cloud computing seem so new?  That, I think, is an artifact
  4375. of history.  In the 1970s and early 1980s, computers were expensive and
  4376. timesharing was popular. In the 1990s and early 2000s, computers became
  4377. increasingly cheaper, and running one&apos;s own became enormously popular.
  4378. Timesharing faded away as people bought and ran their own computers.
  4379. Now the pendulum is swinging back, not driven so much by the cost
  4380. of computers themselves, but the costs of datacentres to house them.
  4381. A few years ago, &lt;A HREF=&quot;http://www.amazon.com&quot;&gt;Amazon Inc.&lt;/A&gt; saw a
  4382. business opportunity in making virtual machines available for rental:
  4383. it was building grid computers (and datacentres to house them) for its
  4384. own operations anyway; why not rent out some of those computing resources
  4385. to other firms? In so doing, Amazon developed an important new line of
  4386. business.  At the same time, a huge number of new internet firms arose,
  4387. such as Netflix, whose operations are dominantly or exclusively that of
  4388. providing various computer-related services over the internet, and it made
  4389. a great deal of sense for such firms to use Amazon&apos;s service.  After all,
  4390. when a company&apos;s operations are primarily or exclusively serving customers
  4391. on the internet, why not make use of computing resources that are already on
  4392. the internet, rather than build private datacentres (which takes time,
  4393. money and expertise)?  These new internet firms, with lines of business
  4394. that were not even possible a decade or two ago, and Amazon&apos;s service,
  4395. also only a few years old, have lent their sheen of newness to the notion
  4396. of &quot;cloud computing&quot; itself, making it appear fresh, inventive, novel.
  4397. But is it?  The name is new, yes. But in truth, the concept is almost
  4398. as old as commercial computing itself: it has merely been reinvented for
  4399. the internet.
  4400.  
  4401. &lt;p&gt; Of course, the computing field, because of its inventiveness, high rate
  4402. of change and increasing social profile, is rather at risk of falling into
  4403. trendiness, and cloud computing certainly has become a significant trend.
  4404. The danger of trendiness is that some will adopt cloud computing not on
  4405. its own merits, but solely because it seems to be the latest tech tsunami:
  4406. they want to ride the wave, not be swamped by it.  But cloud computing
  4407. is complex, with many pros and cons; it is certainly a legitimate choice,
  4408. as was timesharing before it, but it is not necessarily the best thing for
  4409. everyone.  It&apos;s easier to see this, I think, if we look beyond the name,
  4410. beyond the trend, and see that the &quot;rent or buy&quot; question for computing
  4411. has been with us for decades, and the decision between renting virtual
  4412. machines and buying physical ones has often been complex, a balance of
  4413. risks, opportunities, and resources.  For an internet firm whose customers
  4414. are exclusively on the internet, renting one&apos;s computing assets on the
  4415. internet may make a great deal of sense.  For other firms, it may not make
  4416. sense at all.  Deciding which is true for one&apos;s own firm takes wisdom and
  4417. prudence; a healthy dose of historical perspective is unlikely to hurt,
  4418. and may help cut through the hype.
  4419. </description>
  4420.  </item>
  4421.  <item>
  4422.    <title>Handling Unsolicited Commercial Email</title>
  4423.    <pubDate>Tue, 23 Apr 2013 12:56:00 -0400</pubDate>
  4424.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2013/04/23#uce</link>
  4425.    <category>/misc</category>
  4426.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/uce</guid>
  4427.    <description>
  4428. &lt;p&gt; My email address is all over the web: at the time of writing this, &lt;A
  4429. HREF=&quot;http://www.google.ca/search?q=%22jdd%40cs.toronto.edu%22&quot;&gt;a search
  4430. on google&lt;/A&gt; for my email address produces about 15,800 results.  So
  4431. anyone who wants to find my email address can do so easily.
  4432.  
  4433. Many people or companies who want to sell me something send me email
  4434. out of the blue.  I get a great deal of such unsolicited commercial
  4435. email, too much to read or pay adequate attention to.  I simply delete
  4436. them. Unfortunately, many sources of such email persist.  So for some time
  4437. now, I&apos;ve elicited the help of technology.  I process my incoming email using
  4438. &lt;A HREF=&quot;http://www.procmail.org&quot;&gt;procmail&lt;/A&gt;, a powerful piece of software
  4439. that lets me script what happens to my email.  When I receive unsolicited
  4440. commercial email, if it is from a vendor or organization I don&apos;t have a
  4441. relationship with, I will often add a procmail rule to discard, unseen, all
  4442. future email messages from that vendor.  I&apos;ve got about 400 organizations
  4443. (mostly vendors) in my discard list so far, and the list slowly grows.
  4444. Am I still getting unsolicited commercial email from these sources?  I am,
  4445. but I am not seeing it.  It&apos;s the same effect, really, as manual deletion
  4446. (i.e.  the message is deleted, unread), but it&apos;s easier for me, because
  4447. I am not interrupted.  But of course I think it would be better still if
  4448. the email were not sent at all.
  4449.  
  4450. &lt;p&gt; If you are a vendor with whom I do not have a pre-existing relationship,
  4451. and you want to send me email introducing your products, please don&apos;t.
  4452. &lt;A HREF=&quot;blog.cgi/2011/12/15#coldcalls&quot;&gt;I do not accept cold salescalls
  4453. either.&lt;/A&gt; Instead, advertise effectively on the web, so that if I am
  4454. looking for a product like yours, I can find you.  If you must contact
  4455. me directly, send me something by postal mail, where, unlike email, the
  4456. communication does not have an interruptive aspect.
  4457. </description>
  4458.  </item>
  4459.  <item>
  4460.    <title>A closer look at the University of Toronto&apos;s international ranking in Computer Science.</title>
  4461.    <pubDate>Thu, 29 Nov 2012 00:00:00 -0500</pubDate>
  4462.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2012/11/29#rankings</link>
  4463.    <category>/misc</category>
  4464.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/rankings</guid>
  4465.    <description>
  4466. &lt;p&gt;International rankings of universities seem to be all the rage
  4467. these days.  The interest seems to be fed by three rankings of
  4468. particular prominence that have emerged in the past decade. These
  4469. are &lt;A HREF=&quot;http://en.sjtu.edu.cn&quot;&gt;Shanghai Jiao Tong University&lt;/A&gt;&apos;s
  4470. &lt;A HREF=&quot;http://www.shanghairanking.com/&quot;&gt;
  4471. Academic Ranking of World Universities&lt;/A&gt; (sometimes known
  4472. as AWRU, or simply as the &quot;Shanghai Ranking&quot;), &lt;A HREF=&quot;http://www.qs.com&quot;&gt;Quacquarelli
  4473. Symonds&lt;/A&gt;&apos; &lt;A HREF=&quot;http://www.topuniversities.com&quot;&gt;QS
  4474. World University Rankings&lt;/A&gt;, and the &lt;A
  4475. HREF=&quot;http://www.timeshighereducation.co.uk/world-university-rankings/&quot;&gt;Times
  4476. Higher Education World University Rankings&lt;/A&gt;.   Part of the attractiveness
  4477. of these rankings is that they can become a way of &quot;keeping score&quot;, of
  4478. seeing how one institution does in comparison to others.
  4479.  
  4480. &lt;p&gt;My employer, the &lt;A HREF=&quot;http://www.utoronto.ca&quot;&gt;University of
  4481. Toronto&lt;/A&gt;, does quite well in these rankings, particularly my department,
  4482. &lt;A HREF=&quot;http://www.cs.toronto.edu&quot;&gt;Computer Science&lt;/A&gt;.  The subject
  4483. area of Computer Science is not ranked separately in the &lt;A HREF=&quot;&quot;&gt;Times
  4484. Higher Education World University Rankings&lt;/A&gt; (it&apos;s bundled together with
  4485. Engineering), but in the other two, Toronto has consistently ranked in the
  4486. top ten in the world each year in Computer Science, with only one exception.
  4487.  
  4488. &lt;p&gt;This exception is recent, however, and worth a closer look.  In the &lt;A
  4489. HREF=&quot;http://www.topuniversities.com&quot;&gt;QS World University Rankings&lt;/A&gt;
  4490. for Computer Science and Information Systems, Toronto dropped from &lt;A
  4491. HREF=&quot;http://www.topuniversities.com/university-rankings/world-university-rankings/2011/subject-rankings/technology/computer-science-information-systems&quot;&gt;10th
  4492. in 2011&lt;/A&gt; to &lt;A
  4493. HREF=&quot;http://www.topuniversities.com/university-rankings/world-university-rankings/2012/subject-rankings/technology/computer-science-information-systems&quot;&gt;15th
  4494. in 2012&lt;/A&gt;.  This big drop immediately raises all sorts of questions: has
  4495. the quality of Toronto&apos;s Computer Science programme suddenly plummetted?
  4496. Has the quality of Computer Science programmes at other universities suddenly
  4497. soared?  Or has the QS World University Rankings changed its methodology?
  4498.  
  4499. &lt;p&gt;To answer this question, let&apos;s look at how other universities have
  4500. changed from 2011 to 2012 on this ranking.  Many (MIT, Stanford, Berkeley,
  4501. Harvard, Oxford, Cornell, and others) stayed where they were.  Others dropped
  4502. precipitously: Cambridge University dropped from 3rd to 7th, UCLA from 8th to
  4503. 12th, and Caltech plummetted from 7th to 27th.  Some other universities
  4504. went up: Carnegie Mellon University (CMU) went from 9th to 3rd, ETH Zurich
  4505. from 11th to 8th, the National University of Singapore (NUS) from 12th
  4506. to 9th, and the Hong Kong University of Science and Technology (HKUST)
  4507. soared from 26th to 13th.  Surely these curious and significant changes
  4508. reflect a methodology change?  But what?
  4509.  
  4510. &lt;p&gt;The QS university rankings website, in the Methodology section, Academic subsection, reveals &lt;A HREF=&quot;http://www.iu.qs.com/university-rankings/subject-tables/?&quot;&gt;something of interest:&lt;/A&gt;
  4511. &lt;PRE&gt;
  4512. NEW FOR 2012 - Direct Subject Responses
  4513.  
  4514. Until 2010, the survey could only infer specific opinion on
  4515. subject strength by aggregating the broad faculty area opinions
  4516. of academics from a specific discipline. From the 2011 survey
  4517. additional questions have been asked to gather specific opinion
  4518. in the respondent&apos;s own narrow field of expertise. These responses
  4519. are given a greater emphasis from 2012.
  4520. &lt;/PRE&gt;
  4521. To understand this change, it needs to be recognized that the QS rankings
  4522. rely highly on the opinions of academics.
  4523. A large number of academics around the world are
  4524. surveyed: the QS rankings website indicates that in 2012, &lt;A
  4525. HREF=&quot;http://www.iu.qs.com/university-rankings/academic-survey-responses/&quot;&gt;46079
  4526. academic responses were received, of which 7.5% addressed Computer
  4527. Science.&quot; &lt;/A&gt;  The seemingly modest change made in 2012, to weigh more
  4528. heavily the opinions of academics in a field about their own field, given
  4529. its impact on the 2012 results for Computer Science, leads one to wonder
  4530. about the regional distribution of academics in Computer Science in
  4531. comparison to academics in other disciplines.  One significant factor may
  4532. be China.
  4533.  
  4534. &lt;p&gt;&lt;A
  4535. HREF=&quot;http://fusion.grids.cn/career/attachments/china-SIGCSE2010.pdf&quot;&gt;In
  4536. 1999, courses in the fundamentals of computer science became required
  4537. in most Chinese universities, and by the end of 2007, China had nearly a
  4538. million undergraduates studying Computer Science&lt;/A&gt;.  While QS rankings
  4539. does not indicate regional distribution by discipline for the academics
  4540. whose opinions it consults, the surge in the number of Chinese computer
  4541. scientists worldwide in the past decade almost certainly must have an
  4542. effect on the regional distribution of academics in Computer Science
  4543. as compared to other disciplines.  As such, is it any surprise to see
  4544. world universities prominent in China that possess strong Computer Science
  4545. programmes (such as HKUST and NUS) climb significantly in the rankings,
  4546. and others less prominent in China plummet?  But if a world ranking of
  4547. universities is so affected by regional shifts in those whose opinion
  4548. is being solicited, how reliable is it as an objective gage of the real
  4549. quality of a given university?
  4550.  
  4551. &lt;P&gt;Perhaps a more reliable gage of quality can be found
  4552. in the Shanghai ranking, which is not opinion-based, but &lt;A
  4553. HREF=&quot;http://www.shanghairanking.com/aboutarwu.html&quot;&gt;relies on concrete
  4554. indicators and metrics&lt;/A&gt;.  On the Shanghai ranking, the University of
  4555. Toronto consistently ranks 10th in the world in Computer Science in 2010,
  4556. 2011, and 2012.  But what does this mean, concretely?
  4557.  
  4558. &lt;P&gt;To answer these questions, we need to grapple with an
  4559. important fact: in Computer Science, the US dominates.  As a nation, &lt;A
  4560. HREF=&quot;http://books.google.ca/books?id=jv1m50j4ZAMC&quot;&gt;the
  4561. US has been enormously supportive of Computer Science ever since the field
  4562. first existed, and as a result, it has become pre-eminent in computing&lt;/A&gt;.
  4563. Nine of the top ten schools in the Shanghai ranking, and twenty of the
  4564. top twenty-five, are in the US.  For the University of Toronto to
  4565. be one of the handful of universities outside the US to break into
  4566. the top twenty-five, and the only one to break into the top ten,
  4567. is a significant accomplishment.  A chart is illustrative:  &lt;p&gt;&lt;IMG
  4568. SRC=&quot;http://www.cs.toronto.edu/~jdd/blog/misc/rankings/shanghai-cs-nonus-2012.jpg&quot; width=800 height=600&gt;
  4569.  
  4570. &lt;p&gt;Of course, the University of Toronto is in Canada, so a comparison
  4571. to other schools in Canada is also illustrative.  For Computer Science,
  4572. on the Shanghai ranking, there seems to be no close Canadian rival.
  4573. In 2012, UBC comes closest, being a only a few points short of breaking
  4574. into the top 25, but all other Canadian schools rank well back:
  4575. &lt;p&gt;&lt;IMG SRC=&quot;http://www.cs.toronto.edu/~jdd/blog/misc/rankings/shanghai-cs-can-2012.jpg&quot; width=800 height=600&gt;  
  4576.  
  4577. &lt;P&gt;Even compared to other disciplines
  4578. that have Shanghai rankings (only science, social science, and related
  4579. disciplines seem to be ranked), Toronto&apos;s pre-eminence in Computer Science
  4580. in Canada is striking:
  4581. &lt;p&gt;&lt;IMG SRC=&quot;http://www.cs.toronto.edu/~jdd/blog/misc/rankings/shanghai-can-2012.jpg&quot; width=800 height=600&gt;
  4582.  
  4583. &lt;p&gt;From a score-keeping perspective, I think we can conclude that the
  4584. University of Toronto is doing very well in Computer Science with respect
  4585. to other universities in Canada, and it is one of the few non-US
  4586. schools that can keep up with the US in this field.
  4587.  
  4588. &lt;p&gt;But all this needs to be put into perspective.  After all, rankings
  4589. are not a full picture, they&apos;re aggregations of metrics
  4590. of varying value, they represent a formulaic approach to something
  4591. (university education) that cannot always be so conveniently summarized,
  4592. and they reflect methodologies chosen by the producers of the rankings,
  4593. methodologies that may not always best reflect objective quality.
  4594. Of course, if the University of Toronto were to climb to fifth, I&apos;d be
  4595. pleased, and if it were to drop to fifteenth, I&apos;d be disappointed: surely
  4596. the score-keeper in me can be allowed this much.  But in the overall scheme
  4597. of things, what matters most for Computer Science at Toronto is not our
  4598. score on a ranking system, but the objective quality of our programme,
  4599. the learning outcomes of our students, and the impact of our research, and
  4600. these things, not our score on rankings, must always remain our top priorities.
  4601. </description>
  4602.  </item>
  4603.  <item>
  4604.    <title>Intel desktop CPU price-performance: Hyperthreading not helping?</title>
  4605.    <pubDate>Wed, 22 Aug 2012 14:07:00 -0400</pubDate>
  4606.    <link>http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2012/08/22#intelcpus</link>
  4607.    <category>/it</category>
  4608.    <guid isPermaLink="false">http://www.cs.toronto.edu/%252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/intelcpus</guid>
  4609.    <description>Typically, CPU prices follow performance.  Faster CPUs command higher
  4610. prices; slower CPUs are available for less.  Recent Intel desktop CPUs
  4611. continue to show this general pattern, but there appears to be more to the
  4612. story than usual.
  4613.  
  4614. &lt;P&gt; At first glance, everything seems to be what you would expect. Using current
  4615. pricing in US$ at time of writing from &lt;A HREF=&quot;http://www.newegg.com&quot;&gt;newegg.com&lt;/A&gt;, we get:
  4616. &lt;table border=1&gt;
  4617. &lt;tr&gt;&lt;th&gt;Processor&lt;/th&gt; &lt;th&gt;PassMark&lt;/th&gt; &lt;th&gt;Price&lt;/th&gt; &lt;th&gt;PassMark/$&lt;/th&gt; &lt;th&gt;Price-Performance vs G640&lt;/th&gt;&lt;/tr&gt;
  4618. &lt;tr&gt;&lt;td&gt;Pentium G640&lt;/td&gt; &lt;td&gt;2893&lt;/td&gt;  &lt;td&gt;$79&lt;/td&gt;  &lt;td&gt;36.6&lt;/td&gt;        &lt;td&gt; 100%&lt;/td&gt;&lt;/tr&gt;
  4619. &lt;tr&gt;&lt;td&gt;i3-2120&lt;/td&gt;      &lt;td&gt;4222&lt;/td&gt;  &lt;td&gt;$125&lt;/td&gt; &lt;td&gt;33.8&lt;/td&gt;        &lt;td&gt; 92.2%&lt;/td&gt;&lt;/tr&gt;
  4620. &lt;tr&gt;&lt;td&gt;i5-3570&lt;/td&gt;      &lt;td&gt;7684&lt;/td&gt;  &lt;td&gt;$215&lt;/td&gt; &lt;td&gt;35.7&lt;/td&gt;        &lt;td&gt; 97.6%&lt;/td&gt;&lt;/tr&gt;
  4621. &lt;tr&gt;&lt;td&gt;i7-3770&lt;/td&gt;      &lt;td&gt;10359&lt;/td&gt; &lt;td&gt;$310&lt;/td&gt; &lt;td&gt;33.4&lt;/td&gt;        &lt;td&gt; 91.3%&lt;/td&gt;&lt;/tr&gt;
  4622. &lt;/table&gt;
  4623. The &lt;A HREF=&quot;http://www.cpubenchmark.net/&quot;&gt;PassMark (http://www.cpubenchmark.net/&lt;/A&gt;) to
  4624. dollar ratio is pretty consistent across all these processors, roughly 35 &amp;plusmn; 2.
  4625.  
  4626. &lt;P&gt; But what happens if we look at a more real-life benchmark?  Consider
  4627. SPEC CPU 2006 Integer (CINT2006) Baseline.  For each CPU, I used the CINT2006 Baseline results from
  4628. the most recently reported Intel reference system, as reported on &lt;A HREF=&quot;http://spec.org&quot;&gt;spec.org&lt;/A&gt;.  In the case of the G640, no
  4629. Intel reference system was reported, so I used the results for
  4630. a Fujitsu Primergy TX140 S1p.
  4631. &lt;table border=1&gt;
  4632. &lt;tr&gt;&lt;th&gt;Processor&lt;/th&gt; &lt;th&gt;CINT2006 Base&lt;/th&gt; &lt;th&gt;Price&lt;/th&gt; &lt;th&gt;CINT/$&lt;/th&gt; &lt;th&gt;Price-Performance vs G640&lt;/th&gt;&lt;/tr&gt;
  4633. &lt;tr&gt;&lt;td&gt;Pentium G640&lt;/td&gt; &lt;td&gt;34.4&lt;/td&gt;  &lt;td&gt;$79&lt;/td&gt;  &lt;td&gt;0.44&lt;/td&gt;        &lt;td&gt; 100%&lt;/td&gt;&lt;/tr&gt;
  4634. &lt;tr&gt;&lt;td&gt;i3-2120&lt;/td&gt;      &lt;td&gt;36.9&lt;/td&gt;  &lt;td&gt;$125&lt;/td&gt; &lt;td&gt;0.30&lt;/td&gt;        &lt;td&gt; 67.8%&lt;/td&gt;&lt;/tr&gt;
  4635. &lt;tr&gt;&lt;td&gt;i5-3570&lt;/td&gt;      &lt;td&gt;48.5&lt;/td&gt;  &lt;td&gt;$215&lt;/td&gt; &lt;td&gt;0.23&lt;/td&gt;        &lt;td&gt; 51.8%&lt;/td&gt;&lt;/tr&gt;
  4636. &lt;tr&gt;&lt;td&gt;i7-3770&lt;/td&gt;      &lt;td&gt;50.5&lt;/td&gt;  &lt;td&gt;$310&lt;/td&gt; &lt;td&gt;0.16&lt;/td&gt;        &lt;td&gt; 37.4%&lt;/td&gt;&lt;/tr&gt;
  4637. &lt;/table&gt;
  4638. When looking at CINT2006 Baseline, we see the price-performance ratio drop off dramatically as
  4639. the processor price increases.  We would expect this from the i3 to the i5,
  4640. since SPEC cpu int is a single job benchmark and the i3 to the i5 represents
  4641. a transition from two to four cores, but it&apos;s curious to see the dropoff
  4642. in the price-performance ratio between the G640 and the i3 (both dual-core
  4643. CPUs), and the i5 and the i7 (both quad-core CPUs).  What might be going on?
  4644.  
  4645. &lt;P&gt; A look at hyperthreading may provide some answers.  Intel hyperthreading is a feature
  4646. of some Intel CPUs that allow each physical core to represent itself to
  4647. the OS as two different &quot;cores&quot;.  If those two &quot;cores&quot; simultaneously run
  4648. code that happens to use different parts of the physical core, they can
  4649. proceed in parallel.  If not, one of the &quot;cores&quot; will block.  The i3 and
  4650. i7 CPUs offer hyperthreading, the Pentium G and i5 do not.  It turns out
  4651. that the PassMark benchmark sees significant speedups when hyperthreading
  4652. is turned on.  SPEC CINT2006, and many ordinary applications, do not.
  4653.  
  4654. &lt;P&gt; What about SPEC CINT2006 Rate Baseline, then?  The SPEC CPU Rate benchmarks measure
  4655. throughput, not just single-job performance, so maybe hyperthreading helps
  4656. more here?  Let&apos;s see:
  4657. &lt;table border=1&gt;
  4658. &lt;tr&gt;&lt;th&gt;Processor&lt;/th&gt; &lt;th&gt;CINT2006 Rate Base&lt;/th&gt; &lt;th&gt;Price&lt;/th&gt; &lt;th&gt;Rate Base/$&lt;/th&gt;&lt;th&gt;Price-Performance vs G640&lt;/th&gt;&lt;/tr&gt;
  4659. &lt;tr&gt;&lt;td&gt;Pentium G640&lt;/td&gt; &lt;td&gt;61.7&lt;/td&gt;  &lt;td&gt;$79&lt;/td&gt;  &lt;td&gt;0.78&lt;/td&gt; &lt;td&gt; 100%&lt;/td&gt; &lt;/tr&gt;
  4660. &lt;tr&gt;&lt;td&gt;i3-2120&lt;/td&gt;      &lt;td&gt;78.8&lt;/td&gt;  &lt;td&gt;$125&lt;/td&gt; &lt;td&gt;0.63&lt;/td&gt; &lt;td&gt; 80.7%&lt;/td&gt; &lt;/tr&gt;
  4661. &lt;tr&gt;&lt;td&gt;i5-3570&lt;/td&gt;      &lt;td&gt;146&lt;/td&gt;  &lt;td&gt;$215&lt;/td&gt;  &lt;td&gt;0.68&lt;/td&gt; &lt;td&gt; 87.0%&lt;/td&gt; &lt;/tr&gt;
  4662. &lt;tr&gt;&lt;td&gt;i7-3770&lt;/td&gt;      &lt;td&gt;177&lt;/td&gt;  &lt;td&gt;$310&lt;/td&gt;  &lt;td&gt;0.57&lt;/td&gt; &lt;td&gt; 73.1%&lt;/td&gt; &lt;/tr&gt;
  4663. &lt;/table&gt;
  4664. If we look at the transition from two to four cores (by comparing the i3 to
  4665. the i5), we now see that the price-performance of the i5 is better than the i3:
  4666. this is no surprise, since we are now measuring throughput, and from the i3
  4667. to the i5, we go from two to four cores.  But there still is a dropoff in
  4668. price-performance between the Pentium G and the i3, and again between the
  4669. i5 and the i7.  It&apos;s not as extreme as before, but it is still significant.
  4670. This suggests that hyperthreading may help with throughput, but not as much
  4671. as the increase in price would suggest.
  4672.  
  4673. &lt;P&gt; What does this mean, then?  It suggests the increase in price from a
  4674. non-hyperthreaded to a hyperthreaded Intel desktop processor may reflect more
  4675. an increase in PassMark performance than an increase in real performance.
  4676. Hyperthreading may have a positive effect, it seems, but typically not as
  4677. much as PassMark suggests.  At present, for best real-world price-performance
  4678. in Intel desktop CPUs, I would consider models without hyperthreading.
  4679. </description>
  4680.  </item>
  4681.  <item>
  4682.    <title>How to avoid being fooled by &quot;phishing&quot; email.</title>
  4683.    <pubDate>Tue, 26 Jun 2012 16:56:00 -0400</pubDate>
  4684.    <link>http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2012/06/26#phishing</link>
  4685.    <category>/it</category>
  4686.    <guid isPermaLink="false">http://www.cs.toronto.edu/%25252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/it/phishing</guid>
  4687.    <description>
  4688. A &quot;phishing&quot; email is an email message that tries to convince you to
  4689. reveal your passwords or other personal details.   Most often, it tries
  4690. to send you to a website that looks like the real thing (e.g. your bank or
  4691. your email provider) but is really a clever duplicate of the real website
  4692. that&apos;s set up by crooks to steal your information.  Often the pretence looks
  4693. authentic.  If you fall for it and give your password or other personal
  4694. details, criminals may steal your identity, clean out your bank account,
  4695. send junk email from your email account, use your online trading account
  4696. to buy some penny stock you never heard of, send email to all the people
  4697. in your address book telling them you&apos;re stranded in a foreign country and
  4698. need them to wire money immediately, or do any number of other bad things.
  4699.  
  4700. &lt;p&gt; But there&apos;s a really easy way to avoid being fooled by phishing messages.
  4701. If you get a message that asks you to confirm or update your account details,
  4702. &lt;em&gt;never, ever go to the website using a link that is in the email message
  4703. itself.&lt;/em&gt; Remember, anyone can send you a message with any sort of
  4704. fraudulent claim, containing any number of links that pretend to go to one
  4705. place, but really go to another.  So if you feel you must check, go to the
  4706. website that you know for sure is the real thing: use your own bookmark
  4707. (or type in the URL yourself), not the link in the message.  
  4708. </description>
  4709.  </item>
  4710.  <item>
  4711.    <title>Dealing with unsolicited salescalls (cold calls).</title>
  4712.    <pubDate>Thu, 15 Dec 2011 15:14:00 -0500</pubDate>
  4713.    <link>http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/2011/12/15#coldcalls</link>
  4714.    <category>/misc</category>
  4715.    <guid isPermaLink="false">http://www.cs.toronto.edu/%2525252525252525252525252525252525252525252525252525252525252525252525252525257Ejdd/blog.cgi/misc/coldcalls</guid>
  4716.    <description>
  4717. &lt;p&gt; For many years, I&apos;ve been plagued by unsolicited salescalls. It&apos;s
  4718. not very hard to find my phone number, and various people (mostly in
  4719. the IT realm) call me up out of the blue hoping to sell me something.
  4720. The interruption is unwelcome, even if the product isn&apos;t.  
  4721.  
  4722. &lt;p&gt; For some years now, my policy is to explain to the caller that I don&apos;t
  4723. accept unsolicited salescalls, sincerely apologize, and end the call.
  4724. Occasionally, I am then asked how I am to be contacted.  I explain that I
  4725. prefer to do the contacting myself: when I have a need, I am not too shy
  4726. to contact likely vendors and make inquiries about their products.
  4727.  
  4728. &lt;p&gt; Occasionally I run into someone who is offended by my unwillingness to
  4729. take their unsolicited salescall.  I do feel more than a little sympathy
  4730. for the salesperson when this happens: I imagine they may think I objected
  4731. to something they did, or to their manner.  The fact is, I handle all
  4732. unsolicited salescalls this way.  As for whether it is intrinsicly
  4733. offensive to reject unsolicited salescalls out of hand, I don&apos;t think
  4734. it is.  Indeed, it is natural for a salesperson to want their salescall,
  4735. even if unsolicited, to be better accepted.  But it is unreasonable for
  4736. any salesperson to expect that unsolicited sales inquiries to strangers
  4737. will always be welcome.  But I do apologize, each time, and in general,
  4738. when I so quickly end telephone conversations with salespersons who call
  4739. me out of the blue.
  4740.  
  4741. &lt;p&gt;Dear reader, if you are a salesperson, and you are tempted to contact
  4742. me to sell me something, please do not call.  Instead, just advertise
  4743. generally (and if you must, send me some mail in the post).  Trust me to
  4744. find you when the need arises.  I frequently do.
  4745. </description>
  4746.  </item>
  4747.  </channel>
  4748. </rss>
  4749.  

If you would like to create a banner that links to this page (i.e. this validation result), do the following:

  1. Download the "valid RSS" banner.

  2. Upload the image to your own server. (This step is important. Please do not link directly to the image on this server.)

  3. Add this HTML to your page (change the image src attribute if necessary):

If you would like to create a text link instead, here is the URL you can use:

http://www.feedvalidator.org/check.cgi?url=http%3A//www.cs.toronto.edu/%257Ejdd/blog.cgi/index.rss

Home · About · News · Docs · Terms

 
Copyright © 2002-9 Sam Ruby, Mark Pilgrim, Joseph Walton, and Phil Ringnalda