🚀 go-pugleaf

This group looks like it needs a new controversy, so...

There have lately been a number of currents running through this newsgroup
that on the surface seem to me to be inter-related in interesting and
perhaps useful ways, if only we can make them more explicit. To begin
with, however abrasive he may be, Gilbert Cockton is somewhat justified
in his attack upon 'strong AI', in the name ( I assume ) of epistemology.
Carl Jung once remarked of Freud that while he was a brilliant man, he was,
after all, only a medical doctor and suffered from the disadvantage of
'not knowing enough' about the philosophy of Mind. I think the same claim
can be made of a number of prominent figures in today's AI horizon. 

Take for example, Marvin Minsky's recent claim in "The Society of Mind" 
that "Mind is what brains do". This effectively reduces one of the greatest
mysteries of human thought to a question which Minsky believes can be
answered completely within his own field. Minsky's position is not only 
arrogant and in violation of one of the very principles Minsky pushes at 
us in  "Society of Mind" (namely, that we should be skeptical of simple 
explanations); it also betrays a terrible ignorance of the nature and history
of the philosophical question "What is Mind?" It seems to me that Minsky
wishes to use AI to accomplish with Epistemology what Skinner tried to
do to Psychology with Behaviorism. In both cases, we see an attempt to put
a discipline of rational inquiry on a 'sound, scientific basis' by simply
eliminating from consideration all of the problems which are not easily
addressed empirically. 

I do not think that it is a coincidence that this same Marvin Minsky is
also largely responsible for having derailed research into neural networks
to the extent that one could fairly say he set the field back at least 10
years. I have read both Minsky and Papert defending themselves from this
charge, and I do not accept their claims. For one thing, Minsky points out
that the claims of the Rosenblatt camp needed to be examined analytically,
and that we were being misled by the success of connectionism in small
laboratory experiments and that there were fundamental problems with
perceptrons which he felt compelled to point out. If Minsky really
believed this, he certainly wouldn't have written "Society of Mind", in
which he makes such bold claims as "Mind is what brains do" without
presenting the slightest bit of analytical support. The only support for
Minsky's K-lines, for example, is their utility in actual attempts to
model memory. As far as I know, Minsky has never taken the time to perform
the same sort of analysis of his own theories that he felt so compelled
to do with Rosenblatt's.

I have also read accounts of Minsky himself admitting that he may have gone 
a bit too far, that he saw himself as trying to counter a connectionist 
hysteria, and that he didn't intend to kill it. This comes, of course,
only after it is apparent that what we now call PDP isn't going to die
an easy death. It's unfortunate that he is put in the position of having
to defend himself for having misled so many people. I for one certainly
don't intend to accuse him of that, not because I believe him for an instant,
but because one really can't blame him if he managed to make fools of so
many. One can only blame the fools.

The fact is that Minsky succeeded because the large majority of AI researchers
wanted Minsky to be right, not because he was. Similarly, most AI researchers
have a vested interest in the possibilities of "Strong AI", and are therefore
unwilling to question it. I am sure, for instance, that many people are
quite happy with the propostion, "Mind is what brains do". Once
one understands it, it seems plausible enough. More importantly, it's very
convenient to a person who wants to go about his business of reproducing
Mind within the context of a machine. For if he were not to believe that
the brain, as a physical system, was a necessary and sufficient condition
for Mind, he might have second thoughts about devoting his life's work to
trying to reproduce (or simulate) Mind on a computer. 

Again I am reminded of Minsky, only this time it is his pooh-poohing of 
brain laterality. Minsky doesn't seem to believe in the difference between
the two hemispheres or, at any rate, he seems to feel that too much has been
made of this difference. He even goes so far in "Society of Mind" to make
the 'bsurd statement that the division of the brain into left and right
hemispheres is just as arbitrary as the division into top and bottom or 
front and back hemispheres. This is pop-logic in an extreme form; it appeals
exclusively to people who want to believe it in the first place, and
it contradicts a very extensive body of well respected *scientific* evidence.
One cannot argue, as Stephen Smoliar has done in Minsky's defense, that
he is cautioning against reading too much into the differences between the
two halves of the brain. If one reads Minsky's statements on brain laterality
closely, he will find him to equivocate quite nicely in this respect.
He never exactly says that there are no differences between the two hemispheres
of the brain, but he clearly leaves the reader with the impression that there
are no _important_ differences. This is exactly what he did with Perceptrons :
he never said they were useless, but he left the definite impression in the
bulk of the AI community that they would never be very useful.

To return to the dictum "Mind is what brains do", we have to ask how this
accounts for the qualitative difference between Mind and Digestion as
in the assertion "Digestion is what stomachs do". Admittedly, as scientists
we can discuss the process of thinking in much the same way as we can discuss
the process of digesting. That is to say, we can observe these processes and
analyze our observations in various ways. The problem arises when we realize
that we can also observe the actions of our brains in another way - from
the inside. One cannot do this with digestion. The by-product of the activities
of the brain is the observer. Who would like to compare this with the
by-products of digestion? Are they not radically different things? Indeed,
we cannot even say that the observer-within-the-brain is a thing at all,
and this is what leads certain people to say that it is nothing.

It's all well and good to talk of self-awareness as an emergent property
of the brain. I suppose that it is, but in saying this one really isn't
saying very much. One can easily imagine instructing a computer to 
answer "Yes" to the question "Are you self-aware?", but only a fool would
want to pretend that this is the same thing as what you and I are experiencing
as self. That my 'selfness' is real is, to me, beyond dispute. That it
is beyond the scope of empirical investigation can be shown in a number of
ways - is the 'red' you see the same color as the 'red' I see? are dogs
self-aware? Is a tree? The fact that we cannot answer these questions derives
from the fact that the observer cannot be observed. This does not mean that
the observer does not exist, as some scientists would have us believe.

It seems to me the absolute absurdity to claim that self is an illusion,
which is essentially what Minsky would have us believe. It certainly provides
an easy answer to the question, "How did _I_ emerge from this purely physical
world?" You didn't. You're just fooling yourself. You aren't, which is to say
_you_ don't really exist, you just think you do. The problem
is, if I don't exist, why should I need to be fooled? What's being misled
by the illusion?

I realize that questions such as this don't matter much to the daily activities
of AI researchers. I have no doubt that AI will make substantial and
significant progress without ever have to broach the question, "What is Mind?",
but I also feel that, eventually, someone with a perspective in both AI
and epistemology is going to have to address it, lest we content ourselves
with the meaningless dictum, "Mind is what brains do."


R.Kohout

#include >standard_disclaimer.h>