From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wupost!darwin.sura.net!gatech!mcnc!ecsgate!lrc.edu!lehman_ds Tue Jan 28 12:15:42 EST 1992
Article 3003 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wupost!darwin.sura.net!gatech!mcnc!ecsgate!lrc.edu!lehman_ds
>From: lehman_ds@lrc.edu
Newsgroups: comp.ai.philosophy
Subject: Re: Intelligence Testing
Message-ID: <1992Jan22.114840.134@lrc.edu>
Date: 22 Jan 92 16:48:40 GMT
References: <1992Jan18.150345.15050@oracorp.com> <6031@skye.ed.ac.uk> <42032@dime.cs.umass.edu>
Organization: Lenoir-Rhyne College, Hickory, NC
Lines: 36

In article <42032@dime.cs.umass.edu>, orourke@unix1.cs.umass.edu (Joseph O'Rourke) writes:
> In article <6031@skye.ed.ac.uk> jeff@aiai.UUCP (Jeff Dalton) writes:
>>
>>Why is it [conversation] such good evidence [for understanding]?  
>>Because it works for people?  So what?
>>[...]
>>I don't see any necessary connection between conversation and
>>understanding. 
> 
> If the conversation were so unrestricted that you could turn it
> into an interrogation of a willing subject, then I think it could
> constitute as strong an indication of the existence of understanding
> as is conceivable.  You could probe deeply on specific topics,
> and ask a series of questions that could only be answered by
> someone who truly understands the topic.  It seems to me this
> would be totally convincing.  In the face of such a conversation,
> you could only entertain the possibility that despite appearances,
> there is no understanding, by using the word "understand" in a 
> new sense, a sense that demands a particular mechanism of understanding 
> with no observable consequences.
   The problem is that the word "understands" implies an unobservable
state.  We can never empirically check to see if a system "understands",
therefore we can not try to seperate two things by this criteria.
    If we can test in some way this abstarct idea of "understands" other
than making a logical loop, then we are faced with the problem of
deciding if the machine "understands", otherwise we must assume the 
system does because of the response we get.  I have said it before, and
I will say it again, with any given input, if the output is the same, 
for all logical purposes, we must treat them as equal.  
   I'll be first to admit this does not nail down the idea that the
machine can undertsand simply because we programed it to give the
right responses, but we have no way to know that it doesn't.  Until
we can prove without a doubt that the machine DOES NOT understand, we
must assume it does.
    Drew Lehman
    Lehman_ds@mike.lrc.edu


