Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!apple!agate!shelby!msi.umn.edu!cs.umn.edu!thornley
From: thornley@cs.umn.edu (David H. Thornley)
Newsgroups: comp.ai.philosophy
Subject: Re: Searle, Strong AI, and Chinese Rooms
Message-ID: <1990Nov19.191925.28285@cs.umn.edu>
Date: 19 Nov 90 19:19:25 GMT
References: <1990Nov15.204949.12075@Solbourne.COM>
Organization: University of Minnesota, Minneapolis - CSCI Dept.
Lines: 92

In article <1990Nov15.204949.12075@Solbourne.COM> vic@corona.Solbourne.COM (Vic Schoenberg) writes:
>
>I suppose it's possible that AI researchers are smarter than philosophers,
>but there are other possibilities. For example, Searle may understand
>the issues differently, or he may impose different criteria on a
>satisfactory reply. In the case of the question of whether passing the
>Turing Test in and of itself assures that a system understands a natural
>language, I think both these factors are involved.
>
Certainly, certainly.  It seems to me that Alan Turing was interested
in testing for the existence of intelligence, while Searle is
interested in the nature of intelligence.  To give a gravitational
analogy, Turing is calculating possible observed planetary orbits
based on Newtonian theory to see if planets might follow them,
while Searle is studying the curvature of space, and why it
happens.

>Recall that the very purpose of the Turing Test is to establish an operational
>test for intelligence, bypassing any attempt to agree on the definition of
>what intelligence is, or what it means to understand a language. With the
>Turing Test, we have a mathematician's attempt to bypass these sticky
>questions of philosophy. It isn't surprising that a philosopher should
>be unamused. To a philosopher of mind, this end run around the main
>issues of the day isn't acceptable. Searle isn't satisfied with an
>operational definition of intelligence because this doesn't address the
>issues of subjectivity, qualia, the problem of other minds, and so
>forth that are central to the human experience and constitute the core
>unsolved problems of this area of philosophic study.
>

Frankly, Alan Turing didn't write his little paper to amuse philosophers.
He was trying to come up with an operational definition that people could
use, if and when anybody declared that a machine was intelligent.

The technique of establishing an operational definition for something
you don't understand is very common.  I've seen it applied to gravity,
memory, and a host of other things.  What makes such a criterion useful
is not its theoretical basis, or an ability to capture all members of
a class, but that it establishes some set of instances (in this case,
hypothetical intelligent computers) that we can observe and reason from.

At the very least, it has the virtue of being somewhat objective.  Consider
this Searle person (how did he get into this discussion? :-).  He keeps
saying that brains think.  How does he establish that?  Has he ever
observed a brain removed from the rest of its body, and determined that
it thinks?  How does he know that anybody but himself thinks, if he
is willing to consider that behavior indicating thought may proceed from
another source?

Speaking personally, I don't know that I have a brain.  I have hard
stuff in my head, which corresponds to descriptions and pictures I
have seen of "skulls."  I am assured from many quarters that skulls
of humans (which class I seem to fall into - consider appearance,
physical capabilities, and the fact that medical techniques based
on humans seem to work on me, not to mention genealogical evidence)
contain brains.  I am further assured that electrodes placed upon
my scalp have detected electrical activity consistent with sleeping
and waking (whether this is from an alleged brain or not I do not
know).  Furthermore, I am told that various parts of science, now
somewhat united as cognitive science, tell me that brains are the
source of various functions which description resembles that which
I experience as "thought."  Therefore, there is one entity which I
know thinks, and I have no more than strongly suggestive evidence
that that entity possesses a brain.

Were I therefore to construct the appropriate thought experiment, I
could conclude that there is a possibility that you cannot build
anything capable of thinking with organic materials, but you need
a computer.  If, on the other hand, I grant that intelligent behavior
indicates intelligence, the thought experiment becomes too much like
Descartes' malevolent demon to be even vaguely plausible.

Therefore, I will start taking Searle more seriously when he provides
some sort of criterion of thought that is not ultimately based on
the Turing Test, or when he gives an understandable difference
between the cognitive powers of a human and the proper computer
programs running on an appropriate machine.  (Searle is correct in
that programs don't think; actually, programs don't do anything.
It is the system running the programs that does something.)
I will take the "causal powers" argument seriously when I find
out what "causal powers" are and why computers (*not* programs,
see above) don't have them.  I will take the "symbol grounding"
argument seriously when somebody shows how to test for it, in
a system-dependent way, in a way not dependent on behavior.
(If you ask me what an apple is, what kind of apples I like,
and to pick an apple out of a fruit basket, how do you know I
am referring to apples, and not thinking I am playing some sort
of chess game or discussing the stock market in some weird code?
The possibility that a computer is doing this plays a major role
in the Scientific American article.)

DHT