From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!wupost!uunet!mcsun!news.funet.fi!sunic!liuida!c89ponga Wed Feb 26 12:53:55 EST 1992
Article 3954 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!wupost!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: <1992Feb23.220148.16403@ida.liu.se>
Date: 23 Feb 92 22:01:48 GMT
References: <1992Feb20.001246.18025@oracorp.com>
Sender: news@ida.liu.se
Organization: CIS Dept, Univ of Linkoping, Sweden
Lines: 68

daryl@oracorp.com writes:

>Pontus Gagge writes (in response to Antun Zirdum):

>> Part of the uninterestingness of the table-cheat lies in that it is 
>> explicitly limited to passing the Turing Test; it has no capacities 
>> for anything else.

>By construction, the table lookup program would respond in a manner
>indistinguishable from a human. Therefore, it has every bit as much
>ability to manage *all* situations as a human would. Of course, given
>the way the table is set up, there are no unforeseen situations,
>because the table covers every possible situation a human can find
>himself in.

It can manage all situations *within the Test* as a human would.
However, it is not flexible enough to be adapted to other situations;
for instance, you could not by any reengineering feat turn it into
a visual pattern recognizer (short of repeating the hypothetical
creation process on any possible pattern, ad nauseam). My objection, 
then, is that albeit the table embodies intelligent human behaviour
in a Test, it does not contain the intelligence in any *accessible*
form. This is a demand we must place on an artificial intelligence
in order to call it interesting. This is why I classify the table as
intelligent (of sorts), but not artificial.

I am thus merely trying to define what I mean by "interesting" intelligence.

>> Another part lies in that it is a mere copy of its creator's
>> intelligence, and not really an *artificial* intelligence.

>> Generally, I would deem any "AI"-solution uninteresting if its
>> development did not give us insight into the mechanisms of
>> intelligence. A bit fuzzy, and perhaps tangential, but this is where
>> *my* interest in AI is mainly centered.

>Nobody is proposing the giant table lookup as a "solution" to AI;
>everyone agrees that it is impossible to build in practice. I thought
>the question at hand was philosophical; whether such a program should
>be considered intelligent, conscious, capable of understanding, etc.

As I have posted previously: I deem it (extremely) unlikely, but not 
impossible, that an AI-solution could be a functionally equivalent 
version of a table-cheat; and thus be incapable of anything outside
the Test. This is, of course, nit-picking; but shows that the
Test is less than perfect, something which to me was not immediately
apparent when first I learned of the Test.

>From your telling of it, the table, in a sense, is the codification of
>the mind of its creator, and so should be as intelligent, conscious,
>etc.

Quite. Within the Test.

>In my opinion, the interesting, even astonishing, thing about a human
>brain is that it manages, in a volume of about a liter, to encode the
>same information that would take 10^(10,000,000,000) bits to encode in
>the straight-forward way (in a table).

I quite agree. Anyone who does not find the human brain awe-inspiring 
has not really thought about it, or is a fool. AI is not a trivial task!
--
/-------------------------+-------- 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.    |
\-------------------------+-----------------------------/


