Xref: utzoo comp.ai.philosophy:1088 comp.ai:9589
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!samsung!umich!sharkey!tygra!dave
From: dave@tygra.Michigan.COM (David Conrad)
Newsgroups: comp.ai.philosophy,comp.ai
Subject: Re: Turing Test, what's the point?
Keywords: AI, computers
Message-ID: <1991Jun28.084151.2272@tygra.Michigan.COM>
Date: 28 Jun 91 08:41:51 GMT
References: <612@ckgp.UUCP>
Organization: CAT-TALK Conferencing System, Detroit, MI
Lines: 78

In article <612@ckgp.UUCP> thomas@ckgp.UUCP (Michael Thomas) writes:
>Hi everyone,
>
>	Now it was my understanding that Turing felt that a computer
> would never be able to "Think" or actually be "Intelligent", and
> dispite [sic] the definitions of these terms, we all understand what is
> meant. (I hope) So it was also my understanding that Turing felt
> that the best a computer might be able to do someday is imitate
> a person. So he devised a test to deturmine [sic] if a computer at some
> point could accomplish this task....
>
> So, is this correct so far??? (I believe it is...)
>

No, it's dead wrong (and illiterate, too).  Your understanding?  Why don't
you just read what Alan Turing wrote?  He certainly never said that a
computer would never be able to think.  What he said was that since we
cannot agree on what the question, "Can a computer think?" means, we don't
have much hope of answering it.  So why ask it?  Why not ask a question
about which we can say something constructive?  A question like, "Can a
computer fool a human in the 'Imitation Game'?"

> 	So, does anyone else feel that prehaps [sic] Turing was incorrect in 
> his analysis of future computer technology (current computer technology)?
>        Does anyone feel that this is still a valid test, which should be
> used in every case to deturmine [sic] if a system is intelligent, as compared
> to humans? 
>        Does anyone believe that this test is in anyway valid in
> deturmining [sic] if a system is intelligent? 
>        

Does anyone actually believe that this test was designed to determine
whether or not a system is 'intelligent'?  I suspect that Turing would be
appalled at this suggestion.  How can we design a test for intelligence
when we cannot say exactly what 'intelligence' is?

If you have a formal definition of 'intelligence' then go ahead and
construct a test to see if a given AI fulfills the definition.  If,
however, you are using the common sense of the word, then we can do no
better than a behavioural test.  In fact, this is what we humans *actually
use*; we decide if something is intelligent by its behaviour.  Does anyone
reading this *know* that I am a human, and not a program running on (say) a
Cray Y-MP?  But then why do you all afford me the same respect (I have been
treated most kindly, thank you) that you would a fellow human?

> Has anyone established a test or heard of one that would be better suited
> for deturmining [sic] if a system is intelligent?
>

Many have suggested 'better' tests, but unfortunately they are rarely
based on 'better' (more exact) definitions of 'intelligent'.

You seem to feel that Turing's test is outdated, but I fail to see why.
It is certainly true that no AI has yet come anywhere near the level of
sophistication required to 'pass' the test, so it is still a challenge
to researchers.  Do you believe that a system which passed the test
would still not be 'intelligent'?  It would certainly be of great use.
You can't think that it would be too specialized.  (I seem to recall
that in another article you indicated that all it would be doing is
imitating human behaviour.  My apologies for not quoting the actual
article.)  As a demonstration of the breadth of a system which could
'pass' a Turing test of indefinite length, it would have to give a 
meaningful, well 'thought' out response to recent newspapers and
magazines and Orwell's _1984_ as input along with the question, "Do you
think the world is going the way of _1984_?"

>-- 
>Thank you,
>Michael Thomas
>(..uunet!ckgp!thomas)

David R. Conrad
dave@michigan.com
-- 
=  CAT-TALK Conferencing Network, Computer Conferencing and File Archive  =
-  1-313-343-0800, 300/1200/2400/9600 baud, 8/N/1. New users use 'new'    - 
=  as a login id.  AVAILABLE VIA PC-PURSUIT!!! (City code "MIDET")        =
   E-MAIL Address: dave@Michigan.COM