Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 beta 3/9/83; site nbs-amrf.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!umcp-cs!nbs-amrf!hopp
From: hopp@nbs-amrf.UUCP (Ted Hopp)
Newsgroups: net.books
Subject: Re: Godel, Escher, Bach
Message-ID: <516@nbs-amrf.UUCP>
Date: Sun, 28-Apr-85 13:52:55 EDT
Article-I.D.: nbs-amrf.516
Posted: Sun Apr 28 13:52:55 1985
Date-Received: Mon, 29-Apr-85 06:32:49 EDT
References: <262@cmu-cs-gandalf.ARPA>, <582@tpvax.fluke.UUCP>
Organization: National Bureau of Standards
Lines: 65

> > I was wondering if anyone knew where to get this book in hardcover.  I
> > have never seen it in this form, but would like a hardcover edition if
> > possible. It is a wonderful book, by the way.  
> I'm not certain, but I don't think it was ever published as a hardcover.
> When I got my copy, I don't remember ever seeing it that way, at least.

I bought a hardcover version at B. Dalton about a year after it came out.
I don't know if it is still being published in a hardcover edition.

> I agree that its the sort of book you'd rather have in hard than soft-bound,
> but two years after reading it, I'm not sure I'd agree that it's all that
> "wonderful". For one thing, it's harder reading than many college
> textbooks, and for another, it struck me that LOTS and LOTS of what he was
> trying to say was just a bunch of fluff and trying too hard for effect.

I agree.  It is very difficult reading.  (Perhaps that is why it got
such good reviews :-).)  Also, it struck me as being "too cute".  For
instance, too many of the examples of recursion (figures, structure of
a chapter, structure of a sentence, etc.) seemed to be put there so the
reader could get warm fuzzies discovering all the variations, with
apparently no expository or other motivation.

I wouldn't even think of comparing this book to a college text.  (C.S.
text, that is; it may be good "art", although I don't think so.)  There
are way too many holes in the presentation, and the lessons to be drawn
from his material are not stated clearly (if at all).  For instance, I
found the theme of the ambiguity (or was it the definition?) of the term
"intelligence" using the behavior of the ant colony particularly weak.
Is he trying to say that mechanical responses (can) give rise to
intelligent behavior (whatever that is), or that intelligent behavior
IS some set of mechanical responses, or what?  It struck me as an
apology for the field of Artifical Intelligence, which I found slightly
offensive.  Being not entirely naive on the subject, I feel he didn't
make the case (whatever it was).  Not that he was necessarily wrong,
but he sins through overstatement and oversimplification.  (Quite an
assertion on my part given the complexity of his book! :-)) He brings
to mind, through his failure to follow it, the advice I once got on
effective presentations:  always state the weaknesses of your case before
someone else does; it lends you credibility.  To be fair, Hofstadter does
state from time to time that there is a lot about AI that we don't know,
but he doesn't acknowledge that what we do know about intelligence may be
wrong.

Another example of the weakness of his presentation is his explanation
of the value of idealized models of reality.  He points out that ideal
models reveal patterns that might otherwise remain hidden.  What he
completely misses is that most of our knowledge actually comes from
examining the DIFFERENCES between behavior predicted by an idealized
model and actual behavior of a system, not by examining the idealized
model itself.  (This thesis was examined by Herbert Simon in the late
50's and the 60's when he started studying notions of causality.  See
Simon's collection of papers, "Models of Discovery".)  Since Hofstadter's
book is almost entirely an examination of formal (idealized) models,
I find it a poor study of intelligence.

As you may have been able to tell by now, I wasn't exactly thrilled
with G.E.B.  It was difficult reading, didn't hang together (self-
reference is too broad a theme for his book the way he wrote it), and
was technically inaccurate (principally through omission).  It was
entertaining in many places, but it was too much like reading flawed
"pop science" ("Gee!  We can do this, too....") to have been actually
enjoyable.
-- 

Ted Hopp	{seismo,umcp-cs}!nbs-amrf!hopp