From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!mips!mips!munnari.oz.au!uunet!mcsun!news.funet.fi!sunic!liuida!c89ponga Thu Feb 20 15:21:08 EST 1992
Article 3771 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!mips!mips!munnari.oz.au!uunet!mcsun!news.funet.fi!sunic!liuida!c89ponga
>From: c89ponga@odalix.ida.liu.se (Pontus Gagge)
Newsgroups: comp.ai.philosophy
Subject: Re: Humongous table-lookup misapprehensions
Message-ID: <1992Feb15.190542.11778@ida.liu.se>
Date: 15 Feb 92 19:05:42 GMT
References: <1992Feb12.002312.19459@ida.liu.se> <1992Feb12.172855.19148@organpipe.uug.arizona.edu> <1992Feb14.015835.22216@ida.liu.se> <1992Feb14.150342.21640@cs.yale.edu>
Sender: news@ida.liu.se
Organization: CIS Dept, Univ of Linkoping, Sweden
Lines: 61

mcdermott-drew@CS.YALE.EDU (Drew McDermott) writes:


>  In article <1992Feb14.015835.22216@ida.liu.se> c89ponga@odalix.ida.liu.se (Pontus Gagge) writes:
>  >bill@NSMA.AriZonA.EdU (Bill Skaggs) writes:
>  >
>  >>In article <1992Feb12.002312.19459@ida.liu.se> 
>  >>c89ponga@odalix.ida.liu.se (Pontus Gagge) writes:

>  >>>Is everybody happy that a DFA exists which passes a Turing Test, and
>  >>>does so in a completely uninteresting manner? Is the Turing Test still
>  >>>a good criterion for intelligence?
>  >>>
>  >
>  >>  I'm not happy.

>I'm not happy either.  In one breath you say a "DFA exists"; in the
>next breath, the story about how it comes to exist makes it obvious
>that it never really could exist:

Being naive with respect to mathematical philosophy, I claim that it 
exists in the realist sense. Sorry, I should have made myself clear.

>  >An infinitely dedicated creator is given the task to create the table
>  >for all conversations lasting less than a century. In order to avoid
>  >certain practical problems we give her a turbo-charged longevity
>  >drug, and put her in a timewarp, where she can happily hack away for 
>  >the requisite >>10E100 years. In a moment, in the twinkling of an eye, :-)
>  >she emerges, hands us the table, which proceeds to pass the Turing Test.

>I'm no fan of the Turing Test, but if a machine could pass it, no one
>would for a moment entertain the hypothesis that such a creator had
>actually visited such a timewarp, etc., etc.  So whatever power it had
>as a criterion it still has after hearing this fantasy.

I will spell out the next step of my problem explicitly, then. Given that
a DFA exists that could cheat the Turing Test (insofar as it would not be
competent for anything but the test itself); there exists the possibility
that functionally equivalent, much smaller, and thus attainable DFA:s exist 
that can cheat the Test.

If any such transformation of the humongous table exists, then it is
possible that a Turing-successful AI actually would not be capable of
anything but passing the Test.

The central questions are
 * the existence of such a transformation;
 * if the transformation exists, whether AI researchers actually could 
   manage to go down such a wildly wrong path.

I find it unlikely that AI researchers actually could delude themselves
into producing the table-cheat idiot AI, but any possibility thereof
weakens the Turing Test somewhat as an operationalist intelligence
criterion.
--
/-------------------------+-------- DISCLAIMER ---------\
| Pontus Gagge            | The views expressed herein  |
| University of Link|ping | are compromises between my  |
|                         | mental subpersonae, and may |
| c89ponga@und.ida.liu.se | be held by none of them.    |
\-------------------------+-----------------------------/


