From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!convex!mips.mitek.com!spssig.spss.com!markrose Thu Apr 30 15:23:11 EDT 1992
Article 5310 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!convex!mips.mitek.com!spssig.spss.com!markrose
>From: markrose@spss.com (Mark Rosenfelder)
Subject: Re: Intelligence, awareness... oh no, back to the Turing Test!
Message-ID: <1992Apr28.185141.29465@spss.com>
Date: Tue, 28 Apr 1992 18:51:41 GMT
References: <1992Apr27.083621.9441@ccu.umanitoba.ca> <1992Apr27.173029.36491@spss.com> <1992Apr28.062159.1931@ccu.umanitoba.ca>
Nntp-Posting-Host: spssrs7.spss.com
Organization: SPSS Inc.
Lines: 51

In article <1992Apr28.062159.1931@ccu.umanitoba.ca> zirdum@ccu.umanitoba.ca 
(Antun Zirdum) writes (quoting me):
>>If you are right, that's bad news for AI.  If our analysis of intelligence
>>is limited to determining by a behavioral test whether it exists, there's
>>no chance of programming it.
>>
>This does not follow! I see everything as behaviour, you would
>be hard pressed to come up with an example that is not behaviour
>that is also accessable to the world. (By this I exclude subjective
>experience - which again I consider to be a special case of behaviour!)

There are two or three issues here; let's not confuse them.

1. Whether intelligence can be detected only by a behavioral test.
2. Whether intelligence is a primitive, or can be broken down into
   its component or contributing characteristics.

My point about "bad news for AI" relates to issue 2; your reply, to issue 1.
You claimed, if I understand you, that intelligence can't be broken down into
smaller components-- "There is nothing deeper than the bottom!", you say.
But if that's so, there's no way to program it, except by chance.

Or perhaps you have a programming approach in mind?  If so, describe it,
and I'll show you how it assumes that intelligence is not a primitive!

>>Again, if humans have an ability to judge whether intelligence exists, they
>>must be using some criteria.  Is it really such a mystery what those
>>criteria are?  How about things like memory, abstract thought, language use,
>>real-world knowledge, creativity, adaptability, problem-solving, model-
>>building, goal-setting, consciousness, judgment?
>>
>Most of these I think that you would agree are behaviour, the rest
>such as consciousness I think that you will argue that these are
>not even related to behaviour? Such things as having four legs can
>be easily refrased as 'behaving as having four legs' - anything else
>that we can say about any physical object can be rephrased in the
>same way - The physical world behaves! Behaviour does not have to
>involve an action - the expenditure of energy, etc.. - behaviour
>can be simply existing.

Again, you're not responding to issue 2.  You seemed to claim that 
"intelligence" could not be broken down into components.  I presented a
list of components.  Do you still maintain your claim?

On issue 1: As you expand the meaning of the word "behavior", your claims 
about behavioral tests mean less and less.  You've simply defined the term
so as to include every form of human knowledge: observable actions,
unobservable mental phenomena, stative predicates.  Well, of course
under that definition a "behavioral" test of intelligence will suffice!
Even Jeff Dalton will be a behaviorist, now-- he will look for the
"behavior" of having a brain!  


