From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!usc!elroy.jpl.nasa.gov!news.claremont.edu!ucivax!orion.oac.uci.edu!nntpsrv Thu Feb 20 15:20:56 EST 1992
Article 3749 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!usc!elroy.jpl.nasa.gov!news.claremont.edu!ucivax!orion.oac.uci.edu!nntpsrv
>From: schoi@teri.bio.uci.edu (Sam "Lord Byron" Choi)
Newsgroups: comp.ai.philosophy
Subject: Look-up tables
Message-ID: <299C3CC6.15601@orion.oac.uci.edu>
Date: 14 Feb 92 22:03:50 GMT
Organization: University of California, Irvine
Lines: 41
Nntp-Posting-Host: teri.bio.uci.edu

I've been restraining myself for a couple months now from posting anything,
hoping to catch up on the discussion before doing so, but I've found this
an impossible task.

It seems that the main argument in favor of a look-up table being able
to defeat the turing test is based on the assumption that the number of
possible conversation is limited.  The argument I have seen articulating
this mentioned a human being's limited life span.

Although it is true that the totality of any human being's conversations
throughout a lifetime is necessarily finite in retrospect, this says nothing
about the possible avenues of conversation at any particular point in
an ongoing conversation.

For instance:  Looking back at past conversation, I might have asked at one 
point the question, "Is four evenly divisible by two?"  After I have asked this
question, it only goes down as one of the many questions I have asked in
my lifetime.  At the instant before I asked the question though, any number
of variations could have been possible.  (e.g. "is five evenlydivisible by
two," "is six evenly divisible by two," "is seven evenly divisible by two" etc).

This possibility clearly defeats any attempt to construct a straight-forward
look-up table.

In order to circumvent this, the programmer could take this structure and
insert an algorithm to translate these words into numbers, calculate, and
then retranlate back into English.  But at this point, the program is no
longer a pure look-up table, includes programming routines which the
programmer "gives" the program a notion of "meaning."  (i.e. a rudimentary
translation of what "divisible" refers to).

Of course, if I were the tester, I would turn around and ask "is a Hostess
Ding-Dong divisible by two?" (admittedly not the most popular formulation
of that question, but clearly understandable by human beings).

The main thesis is that although all the conversations in a lifetime are
finite, the possible conversations are infinite since the question, before it
is presented, is contingent.

Sam Choi
schoi@charity.sas.upenn.edu


