Newsgroups: comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!gatech!news-feed-1.peachnet.edu!news.netins.net!internet.spss.com!markrose
From: markrose@spss.com (Mark Rosenfelder)
Subject: Re: Turing's Playful Games
Message-ID: <D7rpGJ.B5r@spss.com>
Sender: news@spss.com
Organization: SPSS Inc
References: <3k4iub$p8n@oahu.cs.ucla.edu> <3lkf3p$1ph@percy.cs.bham.ac.uk> <D6Gyon.F7r@spss.com> <3nhj4g$i12@percy.cs.bham.ac.uk>
Date: Fri, 28 Apr 1995 23:05:06 GMT
Lines: 70

In article <3nhj4g$i12@percy.cs.bham.ac.uk>,
Aaron Sloman <A.Sloman@cs.bham.ac.uk> wrote:
>markrose@spss.com (Mark Rosenfelder) writes:
>> ...However, Turing hardly described a scientific
>> test (he quite rightly spoke of a "game" instead).  What is an "average
>> interrogator"?
>
>I assumed he meant some sample of people of people who can type,
>chosen randomly, in the manner in which opinion pollsters choose
>their interviewees? (He mentions Gallup polls in his opening
>paragraph.)
[and so on]

I don't doubt that Turing's Imitation Game could be turned into a 
scientific test, by supplying detailed answers to such questions; 
my point was that till this is done, it's not a scientific test.

It's not enough to say "Oh, let's give him the benefit of the doubt";
that's not science.  As Richard Feynman said, "The important thing is
not to fool yourself.  And you are the easiest person to fool."

A couple of comments on specific replies--

>> ....Why just five minutes?
>
>He was a realist? Any time period is arbitrary. The longer the time
>period the harder it becomes for the machine to pass the test. I
>think Turing made a pretty shrewd guess about how difficult progress
>in AI would be, and what might be achieved in 50 years or so.

I'd say he was highly optimistic.  We are already 45 years into the time 
period, and I don't think we yet have a program that could fool the average 
investigator 30% of the time, compared to a human, in 5 minutes.

That "any time period is arbitrary" also strikes me as a bit disingenuous.
The time period for judging if a degree candidate has mastered her subject
is just as "arbitrary"; yet few judges would be content with a mere 5
minute defense.  I don't think it's out of line to ask for a justification
for the time chosen.

>> ...Do we have any
>> assurance that the "average interrogator" can readily distinguish, via
>> a 5-minute teletype conversation, between (say) men and women, children
>> and adults, normal and mentally defective adults, drunk and sober people?
>
>I don't think that's relevant to the test Turing was talking about,
>though I accept that there are different interpretations of what he
>wrote.

I think the question is highly relevant.  Turing wants to conclude *something*
from the ability of human beings to make judgments based on teletype
conversations.  How do we know that humans have any such ability?
Testing the ability to make similar kinds of distinctions would help
answer this question.

>> ..Later in
>> the paper Turing speaks casually of the control ("player B") being omitted;
>> does he not realize that we are now speaking of a very different kind of
>> test?
>
>I am not sure it's a very different kind of test. Or rather, whether
>it is harder to pass that test is an empirical question.

Leaving out the control in other types of scientific test is not generally
considered to be an insignificant change.

Just to liven up this point: we know that some people have conversed with
Eliza and judged it "human".  Do you really maintain that, if asked to
compare a session with Eliza to a session with a human psychotherapist,
they would have been unable to detect any difference?  
