[From Rick Marken (920918.1330)]
Well, I give up. Psychological Science is not going to publish
the "Blind men" paper as a commentary and I'm not going to try to
get it (or any other paper) published in any scholarly journal any
more. I will only publish stuff in the Journal of Living Systems if
it ever exists. The brief comments of the person who reviewed the
"Blind men" paper made it clear to me that attempts to get psychologists
interested in PCT are just futile -- they just don't see anything wrong
with the cuurent state of psychological "science" (as Dag is always
pointing out to me). The fact that their data is junk, that they
have no working models that match the behavior of real organisms,
that their mdoels cannot behave in real environments, that their
concepts of the nature of behavior are based on skewed views of a pervasive
phenomenon that they don't notice or understand (control); all this
is of no interest. They are happy with their statistical tests; I'm
happy with my models. Apparently, there is going to be no dialogue between
conventional psychology and PCT -- other than the one going on right here
no. the test doesn't work backwards; it doesn't have anything to work
backwards to.
If this is true, then the Turing test doesn't seem to "test" much. As
you describe it, the test involves judging whether the behavior I see
seems intelligent or mindlike or whatever. How can that judgement be wrong?
What's the test for? If this is really the Turing test then
it would probably be better described as the Turing experience. I just
recently had this experience in an airport where there were some Coke
cans dancing to the music coming out of a radio. Those cans seemed
like they were really bright; they passed the Turing test. Do you have
any reason to suspect that they didn't?
no. when you go on like this, i see so clearly preston's point that
some people are so locked into the behaviorist/mentalist mindframe
that they can't talk about anything else. (did you ever get her
paper?)
I received and read Preston's paper. It was VERY long. I don't know if
I am locked in the behaviorist/mentalist mindframe. I am locked in
the mindframe of thinking I want to understand human nature. If the Turing
test is actually what you describe then I have no argument with it; I
just don't see what the big deal is. The Turing "test" is apparently nothing
more than the claim that people can have the experience of perceiving
something called "intelligent behavior". Wel, the Marken test goes
way beyond that -- it says that people can experience "moral behavior",
"spiritual behavior", and much much more. I thought Turing (and ai types)
were interested in understanding the mechanisms that produce what is called
"intelligent behavior" -- something which, most folks agree, is produced
by humans. Thus, a measure of the success of the mechanism you build
(or postulate) for producing intelligent behavior is whether it produces
intelligent behavior. And, since the only standard for intelligent behavior
is that which is produced by humans (and described as such), then the
way to test the mechanism is to compare it's behavior (blindly) to that
of a human. If the behavior of the mechanism is indistinguishable from
that of the human, then the mechanism is "artificially intelligent".
If this is not what the Turing Test is about then what is it for? How
is it relevant to ai?
The Turing test (as I understand it) would be just fine if behavior were
what Turing (and everyone else) imagined it to be -- actions and
consequences of those actions. But there is reason to believe that
behavior is controlled perceptual experience. So looking at actions and
their consequences gives a misleading picture of the behavior of the
organism. In order to test whether a hypothesized mechainsm behaves like
the "real" thing you must test to see whether it controls perceptions
like the real thing. Thus, the only problem with the Turing test (from
a PCT perspective) is that it uses the wrong criterion to evaluate the
success of the mechanism proposed as a model of behavior. The Turing
test focused on actions and their consequences -- PCT says look at
controlled perceptions. And it is only possible to look at controlled
perceptions using the test for the controlled variable.
If there is another way to look at this, that's fine with me. I don't
think I want to stay locked in a behaviorist/mentalist mindframe. I
would like to try to see things your way. But, quite frankly, I want
to do science and to me that means observing phenomena and testing
models of those phenomena -- to see if the model acts like the phenomenon.
I admit that I am locked in this mindframe and it seems to me that this is
the mindframe that you believe I should break. It's hard for me to imagine
how I can break it -- even with your help -- because in the small domains
in which I have used this mindframe, it has been most successful. But
please keep trying to convince me. Maybe we could get more concrete and
discuss your perspective in terms of experiments and models.
Chuck Tucker says:
"WAR ON DRUGS" [Re: Marken 920911.0900]
> If there was a
> serious interest in decreasing the use of these special illegal
> substances ... one of the obvious ways to do it is to
> make them legal and dispense them through official means
It sounds like you're saying that there IS a way to control drug use but
the drug war is the wrong way. Maybe I am misreading you but I want to
clarify my post about the drug war anyway. My point was not that the
drug war is a "poor" way to control drug use. My point was simply
that it IS an approach to control drug use (a behavior) -- with the
predictable consequences -- ie. conflict. If you tried to control
drug use by legalizing drugs you would run into the same problem; conflict.
What is important about control of behavior is not the actions that are taken
in an effort to produce the intended result (in this case, lower drug use)
it is simply having the intention that another person will behave in a
particular way -- ie. it is having a reference for a particular perception of
the behavior of another person. If a person is trying to control this percep-
tion (of another's behavior) then he or she will act in whatever way results in
movement of the controlled perception toward the intended level. It is
the intention to control behavior that causes the problem -- not the
actions that are used in an effort to carry out this control. An observer
might judge some of the actions that are used to control a behavior
(like drug use) to be very nice -- like giving the user alternative
drugs or putting the user in a real nice "rancho relaxo" environment to
kick the habit. But as long as it is the reference for drug use of the
controller and not that of the controllee that is being forced as
the one to be adopted by the controllee, then there is conflict and,
possibly, violence.
Re: Stella: I have a copy at work. It looks fine to me. But what I do
can be done just fine in pascal or basic or lotus. But I'll fiddle with
it.
Best regards
Rick
ยทยทยท
on CSGNet. Uncle. Uncle. penni sibun (920914.1300) says:
**************************************************************
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)