Newsgroups: comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!rochester!galileo.cc.rochester.edu!prodigal.psych.rochester.edu!stevens
From: stevens@prodigal.psych.rochester.edu (Greg Stevens)
Subject: Re: Galatea 2.2
Message-ID: <1995Jul3.213230.16747@galileo.cc.rochester.edu>
Sender: news@galileo.cc.rochester.edu
Nntp-Posting-Host: prodigal.psych.rochester.edu
Organization: University of Rochester - Rochester, New York
References: <3t96r5$ght@worm.inch.com>
Date: Mon, 3 Jul 95 21:32:30 GMT
Lines: 35

In <3t96r5$ght@worm.inch.com> <Sonic@inch.com> writes:

>So now that R. Powers' fictional account of contemporary ai has been out 
>a while, Galetea 2.2 published by Farrar Straus Giroux, what is the 
>response from those in the field who have read it? 

It's a wonderful fiction story.  I loved the plot, and much of the dialogue
was witty and insightful.

Or, do you mean, is it predictive or "accurate"?  Up to a certain point
(i.e. Implementations A and B) the results are reasonable, and the
problems that arose with them and the way they are solve are realistic
(i.e. having problems deciding on how many hidden nodes to have, and 
what happens when you have too few or two many, is very real in
connectionism).  There is a slow transition, as in all good fiction, from
those to areas where are only vaguely described and which there is no
implementation to verify the results (i.e. Implementations F, G, and
obviously, H).  

Much of the point of view humerously characterized in Lentz is a nicely
radical characterization of the "Hardline connectionist" AI beliefs --
i.e. in response to the criticism, "Neural nets don't think, they just 
associate" he says (basically), "What is cognition, other than association?",
for example.  All of the arguments brought up against Lentz are very
"true" objections brought up against connectionism, and his responses are
pretty reasonable -- but ultimately whether you believe Lentz comes down
to, as in real life, your personal biases about how you view cognition.

Of course, it makes sense that the viewpoints Powers describes are on
target, seeing as he's a cognitive neuroscientist (I seem to remember 
reading this in an article on this book somewhere).

Greg Stevens
stevens@prodigal.psych.rochester.edu

