X-Message-Number: 8547
Date: Sat, 06 Sep 1997 14:35:59 -0400
From: "John P. Pietrzak" <>
Subject: Why use the Turing Test
References: <>

Mike Perry wrote:

> One thought about the Turing Test is that so far we haven't created a
> program that could pass it, at least at the level of a normal, adult
> human. This suggests such a program must be nontrivial. Whatever it
> *would* have to have, if not "real" intelligence, must be a fair
> imitation. [...] But with intelligence, isn't an "imitation" the
> same as the real thing?

That is the basic rationale behind the test.  To my mind, however,
the ability to mimic a human pattern of responses to queries across
a network connection is not sufficient to encompass the intelligence
of a human.  Initially, many early successes in AI seemed to mimic
intelligence: programs could play chess and checkers, discover
mathematical proofs, diagnose diseases.  Certainly, the humans who
did these things were considered "intelligent".

Unfortunately, each automaton left something to be desired.  There's
something deeper to intelligence, more than just being able to mimic a
particular human ability.  Even when we do have programs regularly
passing a Turing Test, I suspect they will still not be what people
really want.


John

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=8547