Path: utzoo!attcan!uunet!kddlab!trl!rdmei!ptimtc!olivea!apple!usc!sdd.hp.com!caen!sol.ctr.columbia.edu!src.honeywell.com!msi.umn.edu!cs.umn.edu!thornley
From: thornley@cs.umn.edu (David H. Thornley)
Newsgroups: comp.ai.philosophy
Subject: Re: Searle, Strong AI, and Chinese Rooms
Message-ID: <1990Nov30.231103.17041@cs.umn.edu>
Date: 30 Nov 90 23:11:03 GMT
References: <1990Nov15.204949.12075@Solbourne.COM> <1990Nov19.191925.28285@cs.umn.edu> <15798@venera.isi.edu>
Organization: University of Minnesota, Minneapolis - CSCI Dept.
Lines: 88

In article <15798@venera.isi.edu> smoliar@vaxa.isi.edu (Stephen Smoliar) writes:
>In article <1990Nov19.191925.28285@cs.umn.edu> thornley@cs.umn.edu (David H.
>Thornley) writes:
>>
>>Frankly, Alan Turing didn't write his little paper to amuse philosophers.
>>He was trying to come up with an operational definition that people could
>>use, if and when anybody declared that a machine was intelligent.
>>
>[Polite notification that Smoliar wishes to disagree with me]
>
>First of all, for those who do not know this already, Turing's "little paper"
>was published in MIND.  He may not have been interested in amusing
>philosophers, but he certainly considered them the primary audience
>for his observations.  (Remember that Turing spent quite a few hours
>in discussion with Wittgenstein during his Cambridge days, so his thoughts
>about mind date back to before his work on breaking codes or building computing
>machines.)

Guess I should have put the smiley on that comment.
>
>A more important point, however, is that nowhere in this paper does Turing talk
>about operational definitions.  He begins with the question, "Can machines
>think?"  The first thing he does is dismiss this question on the grounds that
>it bites off more than any sensible thinker can chew.  THEN he poses the
>scenario of the "imitation game."  The purpose of posing the scenario is
>to ask whether or not a machine could play it.  He argues that this question
>is more tractable than his original question and then proceeds to discuss how
>one might ultimately build such a machine.
>
Here's how I have read the paper.  First, Turing points out the difficulty
of answering the question, "Can machines think?"  He discusses the male-
female "imitation game," then switches to the human-computer game, suggesting
that a digital computer is a good machine to use.

He then suggests that, in about 2000 AD, machines will exist (with gigabyte
storage) such that they will fool many people much of the time, and also
says that he thinks that, by this time, it will be possible to "speak of
machines thinking without expecting to be contradicted."  My reasoning
from this is that Turing thinks that his test is somewhat connected with
the basic question, "Can machines think?"

Turing then proceeds to discuss nine separate possible objections, or,
as he calls them, "opinions opposed to my own."  He is not clear about
which of his own opinions they are opposed to, but some, particularly
number 4, _The_Argument_From_Consciousness_, do seem to argue that a
machine that can imitate a human sufficiently can be said to think or 
understand or something vaguely like that.

This is why I interpret Turing's paper as supporting the common notion
of the "Turing Test."

>Thus, we are now quite some distance from anything remotely resembling any sort
>of definition (operational or otherwise) for intelligence.  Unfortunately,
>there now seems to be a flood of philosophers of mind who want to read more
>into Turing's paper than he ever intended to write.  The "imitation game" was
>nothing more than an engineering decision to pull thought away from (possibly)
>fruitless speculation and direct it towards something more concrete.  Of
>course, many of us have anecdotes about how some implementation of ELIZA
>managed to play the "imitation game" successfully.  All this means is that
>we have probably now come far enough to think about scenarios more
>sophisticated than Turing's original suggestion.  This seems like
>an excellent thing to do.  Turing introduced the "imitation game"
>to discourage philosophers from idle speculation.  Those philosophers
>now seem to be rushing back to those nebulous words like "think" and
>"intelligence" again.  All this means is that it is time to invent a
>new scenario, more challenging than the imitation game, which can allow
>us to return to more concrete issues again.
>
I don't think the Turing test has been outdated yet; for one thing, I
have not seen anything reliably win the "imitation game" yet, and I do
not expect to see a winner by 2000.  I think the problem of machine
"intelligence" is less tractable than Turing thought.

I do believe that we will eventually produce machines that can pass the
Turing test (in the sense that one believes one's mortgage will be
sold to an out-of-state outfit with bad record-keeping, not in the sense
that one believes in God), and I am sure that people, when interacting
with these machines, will believe they are intelligent, and capable of
thinking and understanding.

I have no anecdotes about Eliza et al. playing the imitation game, just
stories about Eliza being mistaken for human (*not* the same thing).
Nor do I think we need something more challenging than the imitation
game, since we haven't come near making a good player in forty years,
even with storage greatly exceeding the gigawhatever quantities
Turing wrote of.

DHT