BKCAMQUN.RVW   980419

"The Cambridge Quintet", John L. Casti, 1998, 0-201-32828-3,
U$23.00/C$31.95
%A   John L. Casti
%C   1 Jacob Way, Reading, MA   01867-9984
%D   1998
%G   0-201-32828-3
%I   Addison-Wesley Publishing Co.
%O   U$23.00/C$31.95 416-447-5101 fax: 416-443-0948
%P   181 p.
%T   "The Cambridge Quintet"

This book is a work of fiction attempting to teach fact.  The quintet
of characters all existed, but did not all have a connection with
Cambridge, did not all know each other, and did not have the dinner
party described herein.

The premise is this: the British Foreign Office, interested in the
possibilities being claimed for computing machinery, asks C. P. Snow,
physicist, novelist, and (probably most important) civil servant, for
an assessment of the real expectation of producing a "thinking"
machine.  Snow sets up a dinner party to discuss the issue and invites
J. B. S. Haldane, geneticist and science popularizer; Alan Turing,
mathematician and cryptographer; Erwin Schrodinger, physicist and
quantum cat threatener; and Ludwig Wittgenstein, philosopher of
mathematics and language.  To this august group he poses the question:
can machines be made to actually think?

An immediate objection is that the FO wouldn't care.  The question of
what we can get computers to do that might be of use to us is vastly
different from the more philosophical speculation about whether what
we can get computers to do constitutes thought.  Therefore, this
dinner, and book, really serves as a kind of general metaphysical
primer on the various positions regarding the issue of artificial
intelligence.  As well, and in many ways more interestingly, the
debate serves as a pseudobiography of the characteristics and thought
of the guests.

Turing, of course, is used to champion the cause of AI.  He presents
his famous test, and sides firmly with the behaviorists: if it talks
like a mind, and answers like a mind, then it is a mind.  Wittgenstein
is the almost religious opposition: machines can't think because they
can't.  The arguments presented are rather religious as well: thought
transcends analysis and always will, even if symbols can be processed
they still can't mean anything, and universal ideas do not exist. 
Even the most realistic point, that people learn language from social
interactions with other people, could be immediately rebutted by
pointing out that computers, like children, would be programmed by
language using people.  (That this argument is not raised in the book
is odd, but perhaps no odder than the fact that Turing is allowed to
assert that computers would be able to manipulate language with no
requirement being made about the how.)  Schrodinger is a witty
dilettante, not really caring about the outcome but happy to add to or
refine the ideas as they are put forth.  Haldane is used to represent
the pedantic and conservative view, vaguely affronted by the idea of
machine thought and not really understanding the issues.  Haldane's
greatest moment seems to come when he proposes that an entity that
can't get around much isn't much for thinking, either.  (Stephen
Hawking might have a few comments on that.)  Snow is simply used as a
moderator, to interrupt discussions to announce the serving of the
next course, and to restate positions and send them off in different
directions.

The specific concepts that are presented (and, as Casti points out,
many, including the "Chinese Room" and social Darwinism, postdate the
dinner by some years) are not tendered in depth.  The presentation is
a kind of "computing Lite," giving a possibility of recognition, but
nowhere near enough material for full understanding.  I realize that
this is an objection made frequently in regard to "popular" works on
scientific topics, but in this case I really feel that a very slight
increase in length could have increased tutorial utility many times. 
There is, for example, only one illustration of the Turing machine,
where two or three cases might have given a better idea of the general
nature of the device.  In other places, the characters "then [go] on
to explain" such and such, without the explanation ever actually being
given.

The book is easily readable, and quite entertaining at times.  The
material that is presented will hold no surprises for those who have
dealt even peripherally with artificial intelligence, but for those
completely outside, it does present some interesting discussion
starters for the theoretical background.  For our modern fat-free
aesthetes, the dinner is apt to kill you.

copyright Robert M. Slade, 1998   BKCAMQUN.RVW   980419