home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Telecom
/
1996-04-telecom-walnutcreek.iso
/
book.reviews
/
what.computers.cant.do
< prev
next >
Wrap
Text File
|
1995-01-05
|
3KB
|
64 lines
=========================================================================
Date: Mon, 15 Aug 1994 00:21:00 CDT
Reply-To: TK0JUT2@MVS.CSO.NIU.EDU
Sender: CU-DIGEST list <CUDIGEST%UIUCVMD.bitnet@vm42.cso.uiuc.edu>
From: "(Jim Thomas)" <TK0JUT2%NIU.bitnet@vm42.cso.uiuc.edu>
Subject: Cu Digest, #6.72
CONTENTS, #6.72 (Sun, Aug 14, 1994)
File 5--"What Computers Still Can't Do" by Dreyfus (Book Review)
----------------------------------------------------------------------
Date: Thu, 11 Aug 1994 13:33:09 -0600 (MDT)
From: "Rob Slade, Ed. DECrypt & ComNet, VARUG rep, 604-984-4067"
Subject: File 5--"What Computers Still Can't Do" by Dreyfus (Book Review)
The MIT Press
55 Hayward Street
Cambridge, MA 02142-1399
USA
Robert V. Prior, Editor - Computer Science prior@mitvma.mit.edu
Maureen Curtin, Int'l Promo. - curtin@mit.edu
"What Computers Still Can't Do", Dreyfus, 1992, 0-262-54067-3, U$13.95
There are two kinds of classics. In one, an important idea is held
up, explained and illuminated from all sides. This exposure is both
pleasurable and thought-provoking; so much so, that even an opponent
of the central thesis still enjoys the work and may be inspired by it.
The second type of classic takes an important idea and beats it to
death. This is what Dreyfus did with his original work, which is
basically unchanged here. One suspects that his work would not have
produced such animosity had he not taken the tactic of a direct attack
on all of the major artificial intelligence work of the time. (And of
the time since: the introductions to the various reprintings continue
attacks on each succeeding generation of AI.) After a while, even the
most sympathetic reader starts to respond, "So it's limited. So it
doesn't work yet. So what?"
That, however, only applies to the introductions, and part one. Parts
two and three move completely out of the technical realm and into the
philosophical. The second part looks at the philosophical beliefs of
the devotees of AI. "Belief" is the correct term. As Dreyfus points
out at the beginning of part three, committed workers in the field of
artificial intelligence seem to have, consciously or unconsciously, an
almost religious assumption that man, and the brain, is a calculating
device of some kind. Dreyfus points out that these beliefs are
unfounded, in the sense of not being based upon clearly demonstrable
evidence or principles.
However, as one moves into the last part of the book, it becomes
evident that Dreyfus is *not* presenting a critique of artificial
reason or intelligence. He is primarily attacking "cognitive
simulation". Part three presents an alternative view, not of
computing or AI, but of cognition. Unfortunately, this part of the
book, while somewhat interesting, is not as compelling as the negative
parts.
The central thesis of the book, that there are limits to computing and
that we tend to hold unquestioned beliefs about computing power, is
still an important one. The failure, however, to update the book in
terms of current beliefs and aspirations, robs the work of some
impact.