From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!usc!rpi!uwm.edu!linac!att!emory!gwinnett!depsych!rc Tue Jan 21 09:26:58 EST 1992
Article 2868 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!usc!rpi!uwm.edu!linac!att!emory!gwinnett!depsych!rc
>From: rc@depsych.Gwinnett.COM (Richard Carlson)
Newsgroups: comp.ai.philosophy
Subject: Narrative Origins of the Turing Test
Message-ID: <gLZqeB1w164w@depsych.Gwinnett.COM>
Date: 18 Jan 92 13:35:15 GMT
Lines: 96

It has become clear that the Chinese Room is just a "spin" on the
basic Turing Test.  But what is the Turing Test itself?  Is it a
scientific criterion or benchmark of some kind, a "model
experiment" or "critical experiment" or "paradigmatic experiment,"
like Galileo dropping his unequal weights from the Tower of Pizza?
(And whether he actually did or not is irrelevant since the
experiment would have come out right.)  Or Archimedes stepping
into his bathtub?  Or Michelson-Morley?  Or the number of related
species of bugs on the Galapagos Islands?

Let us try to reconstruct the origins of the Turing Test, i.e.,
why Alan Turing proposed it and why it seemed a natural and
common-sensical thing to an English homosexual in the 1930s.
Turing himself, as a gay intellectual in the macho and homophobic
world of the old school tie, which was British science in his day,
had to "pass" a Turing-like test every day of his professional
life.  He had to look like one of the boys, laugh at jokes about
fags and alter the gender of objects in his tales of sexual
conquest. And when he entered the even more macho, 007-like world
of espionage with his Enigma Machine, it was even more necessary
to keep his "cover," a spy story within a spy story, as he broke
the Germans' codes and learned Hitler's and Goering's secrets.

Surely this must have called to his mind various literary
conceits, the mistaken identity so common in Victorian comedy and
melodrama, the crossing of class boundaries in Sheridan's _She
Stoops to Conquer_, the gender difference in plays like _Charlie's
Aunt_ (which probably accounts for the confused suggestion of the
computer simulating female gender in his original proposal of the
Turing Test).  These literary works, built on a Turing-Test-like
theme of "passing" for something one is not must have, consciously
or unconsciously, reminded him of his own daily personal Turing
Test, his facade of heterosexuality.  (I wonder if he ever got
together with Wittgenstein to discuss the "word games" they had to
play with their colleagues.)  It was a simple step to ask himself
something like, well, if a computer could simulate "semantics" and
"understanding," the way I simulate heterosexuality, wouldn't that
prove that it "understood" "understanding" in some sense, just the
way I understand all these straight dudes?

However, a moment's reflection should suffice to suggest the
weakness of Turing's test.  Obviously the fact that he was able to
fake (simulate, mimic, model, whatever) heterosexuality in some
social situations did not in fact "make" him a heterosexual.  All
the while he seemed to be joining his buddies in leering at some
girl, and making them believe he was leering at her, he may have
actually been leering at precisely those buddies who believed he
was co-leering with them at the girl!

I wrote:
]Over the course of the show's development, Vickie does
]learn, so that finally we have an "experienced" robot which can
]pass as a human little girl and the show ends because it has used
]up all possible "situations."  At precisely that time we can ask
]the key question: does she have "semantics" and "understanding"
]... I think that is a way to look at the question.

David Gudeman wrote:
>No, that is exactly the wrong way to look at the question.  You are
>only adding peripheral effects to cloud the issue.  The fact that the
>machine in question appears to the senses exactly like a human does
>not change the philosophical question at all, but it does make it
>harder to take an objective view.
>
>Rather than ask about understanding, why not ask about something
>verifiable?  After Vickie becomes indistinguishable from a human does
>she become flesh and blood?  If a person born in France learns the
>language and customs of Arizona so well that he can convince you he
>was born there, does that mean that he actually was born in Arizona?
>This whole idea --that something must be true just because people can
>be convinced it is true-- is so obviously falacious that I don't see
>how people can seriously propose it.

It was just _because_ Vickie could "pass" for human that these
questions suggested themselves.  Obviously she did not become a
real girl, the way Pinochio became a real boy.  And while a
Frenchman may indeed become an American cowboy -- I did meet a
real live former Frenchman once who approximated that -- his
origins will make for some degree of "difference" between him and
the "real" American cowboys.  You are right in your conclusion, I
believe. The notion that merely because some people are convinced
that something is true it is then true is indeed obviously
falacious. So, then, what good is the Turing Test?  I think it
does tell you something, namely that you understand the process
well enough to model it.  That means that the techniques you have
used, which may not be structurally identical to the processes
being modeled -- the way Alan Turing's leer was not generated by
the same process as his buddies' -- is at least consistent with it
and, around the edges where it breaks down, you can gain further
insight into the processes you are trying to understand.

--
Richard Carlson        |    rc@depsych.gwinnett.COM
Midtown Medical Center |    {rutgers,ogicse,gatech}!emory!gwinnett!depsych!rc
Atlanta, Georgia       |
(404) 881-6877         |


