Turing, Blind men, Hi Isaac

[From Rick Marken (920924.1000)]

Martin Taylor (920923 18:00)--

(Rick Marken 920910.0900)

But I do think that the Turing
Test is an EXCELLENT example of the behavioristic basis of AI (and
cognitive science) etc. Harnad's contribution makes it even clearer -- he
suggests a Total Turning Test meaning its not enough to get a simulation
to answer questions like a real person -- you must also get it to
behave in all ways like a real person -- ie -- brush teeth,
play soccer, build model airplanes, etc. Of course, what is interresting
is that behavior is defined completely superficially -- it is what you
SEE. There is no notion that behavior (including the conversation of the
original Turing Test) is a controlled consequence of simultaneous influences
produced by the actor and the environment.

I think the Turing Test is an excellent model of the basic misconception
about behavior embraced by all the life sciences. According to the
Turing Test (Total version), the only problem with clockwork simluations
of behavior like those done in the 1700s was that they didn't mimic enough
behavior and, perhaps, didn't do it smoothly enough. The Turing Test
shows very clearly that the goal of AI and like minded "sciences" is
truly superficial -- they want to simulate how things "behave: not how
they control.

It was my impression that the Turing Test involved an interaction between the
tester and the testee.

I guess I gave the wrong impression somehow, but, really, I know that
the Turing test is interactive (a dialog) -- note the fifth line of the
first quote above. My point is only that the judgement of "intelligence" is
based on the answers to the tester's queries. These answers are the
behaviors to which I refer. I presume that the tester will take into
account the fact that s/he has asked a question or made a statement and
that the answer relates to it in some way (maybe).

That implies that the tester can apply The Test as
understood by PCT.

Absolutely. The tester can, but doesn't. Similarly, psychologists can view
their manipulations of independent variables as disturbances to hypothetical
controlled variables and watch for lack of effect and/or resistive actions.
But they don't either.

Turing, as I understand him, implied nothing about how
the testee was to arrive at the observed behaviour, or whether the tester
should, could, or would, observe control. So I think the quoted comments
are misguided.

Maybe not clear, but not misguided. Of course the Turing test can be
done as a test for the controlled variable; I've said that over and over
again. My point is that it is NOT done that way (nor is psychological
researh done that way). I'm not criticizing Turing. My point is that the
Turing Test (as it is typically carried out) is behavioristic in the
sense that evaluations of intelligence are based on observed characteristics
of answers to questions (or, in the Total Turing Test, responses to stimuli)
rather than on an understanding that what real intelligent systems (humans)
do is control. I would be thrilled to find evidence that someone, somewhere
had been doing the Turing Test as a test for controlled variables. But I
am not aware that this has been done. Turing described his test in
behavioral terms -- he had no idea that an "intelligent" system might be
a control system and this is certainly understandable; no one else
did either. What I mean by "behavioral" is that the judgment of
"intelligence" was based on whether replies (outputs) to questions seemed
like the kind that would be generated by an intelligent system. All that is
missing from this is the notion that these outputs might be part of an
effort to control perceptual input relative to a reference spec. This may
seem like a small thing to miss -- one which is easily remedied -- and
in a sense it surely is. But it's the whole ball of wax from a PCT
perspective; and in practice, this "easily remedied" little omission has
been actively defended by practicing students of psychology. That is,
psychologists have insisted on keeping this omission (the test for the
controlled variable) omitted from psychological research practice.

But, if you know of examples of the Turing Test or Psychological Research
that was done as a test for controlled perceptual input (other than that
done by PCT afficionados) I would really love to hear about it.

Avery Andrews (920924.1058)--

It seems to me that the problem with the blind men paper is this: academics
just aren't interested in being told how concepts they are used to dealing
with can be reduced to others that they aren't, and abstract arguments that
some other way of looking at things would be better likewise make no

I agree. Some of the best points are sort of hidden away -- like the
main point which is:

When you understand that you are dealing with a control system rather
than an SR, reinforcement or cognitive system then conventional approaches
to research tell you nothing about how that system works; you have to
change the goal and methods of research. The goal is to discover controlled
variables and the method is "the test".

This message is hidden in the last paragraph of the paper.

I just like the paper so damn much (I feel like Pygmalion -- maybe I
should call the paper "Galetea"). I think the "Blind men and the elephant"
image is so beautifully appropriate -- and, of all things, it turns out
that there are THREE ways to describe control (corresponding to the three
major approaches in psychology -- I'd throw psychoanalysis in with the
cognitive approach) just as there were three blind men and the elephant.
So, I admit, the paper's an ego trip -- I really am not interested in
converting people to PCT with it (and, as you note, I will probably
succeed at that with flying colors).

Isaac Kurtzer --

Welcome to the net.

The Player




Richard S. Marken USMail: 10459 Holman Ave
The Aerospace Corporation Los Angeles, CA 90024
E-mail: marken@aero.org
(310) 336-6214 (day)
(310) 474-0313 (evening)