Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!newsxfer.itd.umich.edu!gumby!kzoo!k044477
From: k044477@hobbes.kzoo.edu (Jamie R. McCarthy)
Subject: Re: Reply to Loebner
Message-ID: <1995May2.014329.19470@hobbes.kzoo.edu>
Organization: Kalamazoo College, Kalamazoo MI 49006
References: <3ncug7$igt@nunic.nu.edu> <3o31vj$s5q@theopolis.orl.mmc.com>
Date: Tue, 2 May 1995 01:43:29 GMT
Lines: 47

timd@sun.com writes:

   I'm still very interested in the general consensus on people's
   feelings about A/V and AI. Perhaps the rest of us aren't as
   distinguished in the field as either Loebner or Epstein (two names
   I hadn't heard of before this series of threads, but then, I'm not
   even much of a programmer, so I don't find it surprising), but I'm
   sure many of us have a meaningful opinion.  Perhaps some energy
   should be directed at finding out what other people think about the
   matter?

As a professional programmer with a bachelor's in computer science,
I find it ridiculous that the Turing test should be weighted down by
audio/visual components.  What on earth would this add to the test?
It's obvious what would be taken away -- the core essence of the
problem, which is whether there is a fundamental "humanness" that can
be distinguished from nonhumanness solely via linguistic communication.

The original Turing test, as I understand it, was to distinguish a man
from a woman.  Turing explicitly forbade the judge from seeing either
participant, for obvious reasons, and suggested that handwriting be
excluded as well.  Perhaps Loebner would have the original test
conducted with the man and woman sitting directly in from of the judge.
That would be interesting to Geraldo viewers, but not to me.  A Turing
test where the judge could see images of the participants would be
equally boring.

A Turing test where the participants could see and hear the judge, but
not vice versa, might be interesting twenty years from now when signal
processing is cheap and good, but in today's world it simply piles
impossibly-heavy requirements on the programmers.  And does it add to
the quality of the test?  Not if Helen Keller is disqualified.
Besides, we must try to walk before we try to run.  Obviously, image-
and sound-processing is nontrivial;  understanding human speech is not
simply a matter of hooking up a voice recognizer and a speech
syunthesizer, as one person wrote.  But if a computer can fool a judge
with ASCII text, I'll be damn impressed already;  changing the medium to
audio or video will impress me only marginally more.

(I don't even know if Loebner proposed a one-way or a two-way video and
sound link.  Either, as I've said, goes against the spirit of the test.)

And the requirement that the winning program fool Loebner personally
is unscientific in the extreme and almost beneath comment.
-- 
 Jamie McCarthy       k044477@kzoo.edu       jrm0@aol.com
 http://www.kzoo.edu/~k044477/    I speak only for myself.
