COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	searle[e84,jmc]		Notes for 1984 Sept 4 discussion with John Searle
C00011 00003	There is no law of nature or mathematics that requires that
C00012 ENDMK
C⊗;
searle[e84,jmc]		Notes for 1984 Sept 4 discussion with John Searle

Suppes's advice
Ask for details.  He's strong on generalities.
He's not strong on logic.
Stay away from Wittgenstein's objections to private language.
There's Kripke's new book, but you won't read that by Tuesday.
In general the 1st vs. 3rd person view hasn't been formulated.

My formulation in note to Talmy:
Is epistemology, the theory of knowledge, best considered from the first
person or the third person point of view.  Is the question "How do I know?"
or "How does he (it) know?"?  John Searle favors the first person and
John McCarthy the third person point of view.

Note from Talmy:
I forwarded your memo to me over to John Searle.  He got back to me
saying:

a) He would like this to be a discussion rather than a debate since he
doesn't feel that the issue is so clearly divided into opposing views (or
perhaps that you two occupy completely opposing views).

He would like a format of each person first presenting his perspective
for 15 minutes (with you, perhaps, going first) and then having a
two-way discussion.

b) He felt that epistemology alone might be too narrow a domain, and
suggested expanding the discussion, and its announced title, to "First
Person and Third Person Points of View in Cognitive Science, Philosophy
of Mind, and Epistemology".

c) Given these changes, he felt that your definition of the question
would no longer be accurate and thought any such definition would be
best dispensed with.

Since your message said that any emendations John might make would be
acceptable with you in advance, I'll just go ahead and institute the
above points.  But I wanted to inform you of developments. 

--Len

Points to be made:

1. The issue arose in connection with the discussion about what John calls
"the strong AI thesis" to the effect that machines can really have
beliefs, intentions, etc.  As I understand John to have formulated it, I
regard myself as an adherent of "the strong AI thesis".  I suppose this is
a somewhat exposed position to take, since John is the proprietor of what
"the strong AI thesis" is, he can always add some absurdity to it.
However, if that happens, I'll be able to wriggle out of it somehow.

Getting back to the point, suddenly John accused me of taking a third
person point of view.  Since I had previously decided that what I
was advocating was indeed a "third person point of view", I was
happy to accede to his accusation.  A seven
person panel did not offer an opportunity to pursue the matter.
But here it is.

2. I don't know a lot of history of philosophy, so perhaps I'm
making a mistake in attributing the following to Descartes or
in supposing that he put it more clearly than previous
philosophers.  Anyway, as I understand it, Descartes proposed
to assume as little as possible, and build up a complete
philosophy by pure reason.  I'm not much interested in the
pure reason part, but rather in the idea that the correct
method is to start from as little as possible, preferably
nothing, and build one's knowledge from there using whatever
methods one takes as helpful, e.g. reason, revelation or
observation.  That's what I disagree with.

3. The example I want take analogy from is the foundations of
mathematics.

	This study began perhaps with Cauchy introducing
epsilontics and successively deepened.  Recall the contributions
of Dedekind, Cantor, Frege and Peano.  By the turn of the century,
the subject of foundations had been formulated and the various
philosophical positions began to be developed.  The arguments
concerned what mathematical facts and principles are "obvious"
and what methods of reasoning are valid.  The schools of
intuitionism (Brouwer), logicism (Russell), formalism (Hilbert)
and conventionalism (Poincare) developed.  The arguments raged
on into the twenties, and then metamathematics happened.

	The people I want to mention in this connection are Godel,
Tarski and Gentzen.  The essence of metamathematics is to study
mathematical theories as mathematical objects, using mathematical
methods.  While there were originally some inhibitions about
what mathematical methods to use, the practice and doctrine
evolved of using any mathematical methods that seemed convenient.
Thus one could use classical mathematics to reason about the
set of intuitionist theorems.

4. In order to do AI, it seems to me that we must extend the same
idea to questions that have traditionally been the preserve of
philosophers.  Namely, we need a theory of knowledge seeking
systems, and this surely overlaps the branch of philosophy called
epistemology.  It may not merely overlap it; it may gobble it up.

5. What conjectures can it verify?

6. What if it assumes that the other beings are like itself.

What if it assumes a language and conjectures the truth?  What
can it exclude?

What if it assumes that it has evolved?

Assuming that our world is as it seems to be from common sense and
science, what structure must a program have in order to formulate
and learn and be teachable about the world?

What are the questions of applied epistemology?

Abstract commitment and abstract actions generally.

Chinese room?

	Perhaps this will be surprising, but one of the advantages of
the third person, AI based, way of doing epistemology is that it
allows effective use of information obtained by introspection.
If a researcher forms a hypothesis by introspection that a certain
intellectual result is obtained by a certain process, he is certainly
free to program the mechanism and see if it works and see if its
limitations correspond to those of a human or are worse.
There is no law of nature or mathematics that requires that
when a causal system evolves intelligence, every fact about
the system automatically becomes discoverable and provable
by the intelligence.