In constructing an artificial organism, that is, a system
that exhibits artificial intelligence, one has a choice.
One may stimulate a nonverbal organism-say, a pigeon-
and construct a sensorimotor system, the behavior which
is selected and strengthened by its consequences. Or one
may construct a rule-following system that responds as
directed and changes its behavior as directed....
It is perhaps natural that those who are concerned with
artificial intelligence should choose the second, rule-
following alternative -- resembling a man who drives a
car by making only the moves he is told to make. That
kind of artificial organism could be instructed to behave
in intelligent ways, because specialists in artificial
intelligence are intelligent, and could then be credited
with intelligence.
The first system could, of course, learn to behave
verbally, given the necessary verbal environment,
and it would then resemble the second. The second,
however, would remain forever simply a rule-following
system. A computer may function as either type of
system, but if cognitive scientists have actually
programmed computers to "think creatively" and to
make "scientific discoveries," they have simulated
the first type. And in order to do so, they must have
known a great deal more about it than cognitive science
can tell them.
A few points. Skinner would certainly have agreed with the statement that Carrol quoted from Rosenfield. It is certainly consistent with his own stated views concerning the relationships between neuroscience and a scientific psychology. Also, it should be evident that most traditional forms of AI (artificial intelligence) such as that exemplified by IBM's Deep Blue computer which defeated Kasparov operate on a rule-governed basis while connectionist approaches (represented for example by artificial neural nets) exemplify the types of contingency-shaped systems that Skinner was talking about. Like Skinner I think that if AI is going to take on the task of simulating creative thought it is going to have to take a connectionist approach.
I am also BTW in agrement with Doyle and Chuck concerning the political implications of these different approaches for understanding and simulating human intelligence.
Jim Farmelant
On Wed, 12 Aug 1998 06:49:36 -0800 Doyle Saylor <djsaylor at ix.netcom.com>
writes:
>Hello everyone,
> I wanted to start with a statement I agree with.
>Carrol Cox Aug. 11, 98
>"…The brain, and particularly memory, as described by Rosenfield seems
>to have no room for anything that even vaguely and sloppily could be
>called "structures.""
>
>Doyle
>I don’t know the Israel Rosenfield text. But that is close to both my
>view and the current connectionist understanding of brain neural
>networks as my reading in other areas allows me to grasp these things.
>
>My qualifications concerning neurology is that I read a lot and I’m a
>working class guy. My remarks then aren’t meant to meet the standard
>of
>a claim. However, I do think what Carrol Cox points at merits a
>Marxist
>discussion here. Here is why I think this.
>
>Doyle
>We see that mathematical formalisms such symmetry groups sometimes
>lead
>us to think that the brain has such structures. As Israel Rosenfield
>writes such is not currently thought true of the brain by those who
>believe the connectionist view of the brain, and based upon extensive
>medical research into the brain. The great linguist Chomsky has used
>mathematical formalism to great effect in exploring syntactical
>universalisms of language, and he believes that these formalisms must
>be
>"innate" to the brain. This important theoretical position of Chomsky
>has had a tremendous impact in science by-passing the road blocks
>thrown
>up by behaviorism in psychology and which treated the mind as "black
>box" (an essentially Kantian position). I think of this as a
>scientific
>taboo which was breached right after WWII by both linguists, and a
>nascent computer industry.
>
>Doyle
>The tendency to think of the brain as having mathematical structure
>embedded in it, seems to me to reflect a social and ideological drive
>to
>attribute "god" like qualities to the mind. In other words, an other
>worldly or "transcendental" aspect to the mind. To think of the brain
>as somehow out of this world. This position is dying in our culture
>in
>the sense that taboos about the mind are giving way to a freedom to
>speculate about the mind in a way that for instance Marx himself
>couldn’t have. I really do not believe Hegel has much to contribute
>to
>the understanding of the mind which neural networks give us. But Marx
>was freed by studying Hegel from a mechanistic outlook toward
>economics. For that we ought to be grateful.
>
>Doyle
>Jim Farmelant writes Aug 11, 98:
>"…The question of to what extent a given pattern of behavior is
>contingency-shaped or rule-governed is an empirical one that can only
>be
>settled by an experimental analysis. This view seems IMO consistent
>with the observations of both Chuck and Doyle."
>
>Doyle
>No doubt that the material proof is lacking to decisively reject a
>rule
>governed brain hypothesis. But the obvious is also true from Carrol
>Cox’s citation there is no evidence in the brain so far found of
>extensive "rules" although some "instincts" are present. The idea
>that
>such things as mathematics are external to us, and the mind is
>contingent (or adaptive, and adaptable) has political implications.
>Primarily in our culture, to create a conflict with the aims that
>arise
>in computing to substitute machines for humans. IBM went to a lot of
>trouble to create a computer to defeat the chess player, Gary
>Kasparov.
>And then dropped the issue abruptly after Kasparov was defeated by the
>machine. Clearly IBM and other corporations want to use machines to
>move into brain work. To substitute the machine for the General
>Secretary of the CP as it were. We have to understand this drive in
>Capitalism and why this conflicts with a democratic view of human
>existence. I would say this corporatist idea reflects the God’s point
>of view, that something can know in an absolute sense "truth". As
>socialist we instead rely upon the totality of human actions to "know"
>the world. We don’t have a ‘faith’ that a ruler is anything but
>another
>one of us.
>
>Chuck Grimes writes:
>"…What I am most interested in is dis-placing or eliminating
>structure,
>rules, organization, and other extremely difficult to account for
>appearant aspects of the biological world. I want to displace all
>this
>paraphenalia out into the world and the environment. There is far too
>much concentration on trying to discover hidden design in biological
>processes,…"
>
>Doyle
>I believe this is my opinion too. It is our way of freeing ourselves
>from the kings, and queens that would "rule" us if only we could just
>find them somewhere in the world. We must rely upon ourselves.
>Regards,
>Doyle Saylor
>
_____________________________________________________________________ You don't need to buy Internet access to use free Internet e-mail. Get completely free e-mail from Juno at http://www.juno.com Or call Juno at (800) 654-JUNO [654-5866]