Talk:Hubert Dreyfus's views on artificial intelligence

Two suggested changes in intro to What Computers Can't Do
1. The ISBN for the first edition is listed at the Library of Congress as 0-06-011082-1. My copy of that edition and printing matched their catalog entry.

2. Although wikipedia provides special ISBN tools to link to various manifestations of a given work, ISBNs do not distinguish different editions of a work as well as the standard ID from a national library, e.g. LCCN for U.S. books.

ISBNs confound the physical manifestation and the textual content. This is primarily because the publisher-assigned portion of the ISBN code is interpreted and used variously by different publishers (and even differently by the same publisher over time!) For example, the original publisher of this book, Harper, apparently reserved different blocks of serial numbers within in its ISBN allocation for paperbacks and hard cover bindings, and for different publishing imprints they owned. One can infer this from the fact that a book produced in 1972 with a cloth binding has a higher ISBN assigned than one issued 6 or 7 years later as a paperback reprint.

In contrast, the Library of Congress catalog lists only three entries for this work and each entry refers to a different text: the 1972 first edition, the 1979 second revised edition and 1992 third revised edition. I believe that most wikipedia users will find links to those catalog entries more useful because they reflect revisions of the text rather than the details of a specific print run or binding lot.

Finally, determining the family to which a particular copy of a book belongs can require fairly elaborate bibliographic tools, intuition, and experience. Web tools such as xISBN and thingISBN are only partly successful in imitating algorithmically what experienced human catalogers do. For this book, the xISBN site shows no other ISBNs related to the first edition; The thingISBN site shows 3 others. One might even view their poor performance on this task as a self-referential demonstration of Dreyfus' thesis. ;-) --JevonsBooksStaff 01:35, 21 January 2007 (UTC)

Deletion of the article
This article survived AFD, and yet DXRAW insists on repeatedly deleting it, leaving comments in the log. AFAIK, editors are supposed to improve articles, not destroy them, and discussion should be here, not in the log. In any case, I don't intend to have a revert war here. How do we resolve this? Samfreed 10:43, 4 March 2007 (UTC)
 * You are incorrect, I'm not deleting the article. The summary which i removed was a) way to long and b) not a summary. If you or anybody wants to write a summary then feel free but a chapter by chapter review and riddled with pov does not have place in the article. DXRAW 12:35, 4 March 2007 (UTC)
 * The summary that you removed is actually a VERY, perhaps TOO loyal summary of the book, blow by blow. The POV is all in the original book. I am looking forward to having the time to summarize the summary, as I do take the point that it is too long. HOWEVER, please respect the fact that it is work-in-progress, and part of the progress is to first finish the overall summary so I can sort out what Dreyfus is saying from all his POV stuff. If you can summarize the book better, please do. Otherwise, let me please deal with this non-trivial stuff with due diligence. ALSO, I don't see how your actions respect the debate that happened in the AFD, nor do I see how arguing in the "edit summary" instead of on this discussion page is respectful of the wikipedia guidelines etc. I commend your activism, but please accept that you have been a bit over-assertive here, and revert the deletion. Samfreed 13:32, 4 March 2007 (UTC)
 * The best way would be to copy it to your userspace & work on it there. DXRAW 20:47, 4 March 2007 (UTC)

Human Vs Machine Recall
'PROBLEM: You forget item, but forced to spit something out, you know it's wrong, but when do get forgotten item find your spit-out has correct attributes; EXPERIMENTAL support Dreyfus contentions?'

EXAMPLES: [1] I and colleague talking about the hardware store in mall. Both of us agreed not "Innovation Hardware", but something like that. When we remembered, it was "Restoration Hardware". [2] Mom and I telling Dad nemesis of Mozart was "Scarlatti" (this time it was SO close we even thought we HAD retrieved item). Dad "It's NOT Scarlatti." Right answer "Salieri". [3] a few DOZEN more examples in which attributes of the spit-out (and almost always KNOWN to be wrong answer) are correct; typically consonants, consonant blends/clusters, number of syllables, certain correct syllables -- these are all sound based (phonetics), but see next... [4] Telling colleague during walk to parking garage, "that movie about the kid and the priest and he was faking his dual personality; the title always seemed strange to me, because I couldn't see how it fit movie; can't remember title, but it was something like 'Savage' -- not in sound but in meaning"  Upon recalling movie title it was Primal Fear. Primal and Savage are similar this time SEMANTICALLY

PROBLEM: How can human produce a spit-out that human KNOWS (almost always) to be wrong, yet which has correct attributes (SEE EXAMPLES ABOVE). Even more amazing, sometimes the human KNOWS WHICH ATTRIBUTES ARE CORRECT. For example both I and colleague in [1] knew that the -ation was a correct attribute. How does human produce correct attributes when human has not retrieved the item? I believe this problem COULD NOT EVEN BE FORMULATED for computer simulation/emulation -- nueral nets or no.

QUESTIONS: How can human produce correct attributes when hasn't retrieved forgotten item? How can human even KNOW, at times, which attributes of the forced spit-out are correct when hasn't retrieved item to compare? Am I right that the problem could not even be FORMULATED for computer simulation? Am I right that all of this comprises EXPERIMENTAL evidence in support of Dreyfuss' contentions?

BTW: Excellent article and I did read the book. — Preceding unsigned comment added by 199.196.144.16 (talk)


 * Interesting stuff. I created a shorter title for this section Samfreed 07:05, 16 June 2007 (UTC)

Merge and expansion
I would like to flesh out this article with material from all of his works, including Alchemy and AI and Mind Over Machine, and move the article to Dreyfus' critique of artificial intelligence. I have a draft at User:CharlesGillingham/Drafts/Dreyfus' critique of artificial intelligence. Does anyone object? CharlesGillingham 22:51, 3 December 2007 (UTC)

Comment
We are computers. The brain is a machine, and it can be improved or surpassed. —Preceding unsigned comment added by 84.193.126.141 (talk) 10:34, 22 December 2007 (UTC)


 * Restating your axioms and prejudices does not make them true. This book is about the limits of your world view - reasserting is not progress. Samfreed (talk) 17:20, 23 December 2007 (UTC)


 * "The brain is a machine" may be true if there is a definition of machine that encompasses brain. However, the notion that it: "can be improved" must always depend on what one considers to be better ie that statement is simply an opinion reflecting an attitude towards life and not a statement of fact. LookingGlass (talk) 19:11, 11 November 2012 (UTC)

Merge and expansion, a year later
I finally got around to finishing this draft: User:CharlesGillingham/More/What Computers Can't Do. I'd like to move it over in a couple of days. Let me know if there are problems. CharlesGillingham (talk) 08:33, 2 April 2009 (UTC)


 * ✅ CharlesGillingham (talk) 23:00, 9 August 2009 (UTC)

Inaccurate editorial
"Although clearly AI research has come to agree with Dreyfus"

this is neither clear nor neutral. While the milestones Dreyfus denigrated were not hit in the timelines proposed, progress has in fact been made. Dreyfus' opinions are far from widely accepted in the AI research community at large. — Preceding unsigned comment added by 198.190.171.222 (talk) 21:24, 1 March 2018 (UTC)