COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	EFFECTS OF AI RESEARCH ON PHILOSOPHY - for imitation Sunday Seminar
C00010 ENDMK
C⊗;
EFFECTS OF AI RESEARCH ON PHILOSOPHY - for imitation Sunday Seminar

Quotes from Sloman
sloman[w79,jmc]

	Several times in the last few hundred years, science has taken
a large bite out of philosophy.  Physics took cosmology out of philosophy
in the seventeenth century, Darwin took phylogeny out of philosophy in
the nineteenth, and psychology seceded in the early twentieth century.
Shreds of these subjects remain in philosophy; thus there is still
a "philosophy of extension, succession, space, and number", but the
questions treated are rather tenuous from all but technically philosophical
points of view.
  However, teleological explanations (the purpose of the
ant is to teach us not to be lazy) are dead.

	My view in this paper is that this is about to happen again.
Specifically large parts of epistemology are being taken out
of philosophy proper, re-arranged in emphasis and made technical.
The process is still quite controversial, and this lecture is somewhat
analogous to discussing rabbit stew without too firm a grip on the
rabbit.  I don't really expect that these views will overwhelm
philosophical opinion, although a few philosophers have reached similar
opinions through paths of their own.  The convincing arguments will
have to be successful application of new epistemological ideas in
making artificial intelligence, and this lecture is more likely to
activate the thinking of people with similar inclinations than to
induce people to change firmly held contrary opinions.

	This paper actually was planned for the Sunday Seminar in
Moscow.  Perhaps it was too ambitious, but I hoped to make the
Soviet Marxists even more nervous than they are already and to make
some of the Soviet computer scientists wish they had nerve enough
to attend the Sunday Seminar.  The Marxists are at least subliminally
aware that they destroyed Soviet biology through their dogmatism,
and it had to be re-imported from the West.  Their attack
between 1947 and 1953 on cybernetics and computers also cost
them time they could ill afford.

	Many of them are aware
that even now, they are damaging their country's economics and
sociology.  I don't know whether he has been active in this area
recently, but in the 1960s Victor Glushkov, the very capable, energetic
and political director of the Institute of Cybernetics in Kiev,
wrote articles for Voprosii Filosofii (Problems of Philosophy) in
which he demonstrated that artificial intelligence research was ok
from a Marxist point of view, and it wouldn't advance in directions
threatening Marxism.

	I hoped it would be heartening to the members of the Sunday
Seminar to hear that the philosophical foundations of the state
that was oppressing them were in conflict with that state's goal
of being pre-eminent in technology.  Even though the argument we
can give today for the effects of artificial intelligence research
on philosophy are very incomplete,  there are enough to make those
who support intellectual freedom in science and orthodoxy in other
areas intellectual life fear that they are headed for another disaster.

	Of course, if I'm right some Western philosophers will also
have problems adjusting, but this adjustment can be made at leisure
and without each side trying to get its opponents driven from
their jobs and opportunities to publish.


WHAT AI REQUIRES OF PHILOSOPHY

	
	/a The first thing that artificial intelligence requires of
philosophy is permission to exist.  The idea that a machine can exhibit
those qualities which are called intelligent in a human is repugnant
to many philosophical systems.  Nevertheless, while Dreyfus was mildly
embarassed to be beaten at chess by a program in 1967, a diehard believer
in the impossibility of a demonstration of intelligent performance
can still hope.  However, there isn't at present any second line of
defense; the anti-AI philosophers have not brought themselves to name
any specific level or kind of intellectual performance that they will
claim to show isn't achievable by a computer program.  They have to
take refuge in nebulous arguments that no specific performance would
be a sign of real intelligence.

	This is by no means the worst of it.
The direction taken by AI research itself provides even more problems.

	/a 

epistemology and heuristics

formalizability in logical languages

common sense reasoning

meta-epistemology

what is an intellectual process?  When is one well defined?

Dennett's "Artificial intelligence as philosophy and psychology" refers
to AI having solved Hume's problem of the homunculi and having invented
the frame problem.