COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	OUTLINE OF DEBATE PRESENTATION
C00008 ENDMK
C⊗;
OUTLINE OF DEBATE PRESENTATION


	1. Artificial intelligence is the study of intellectual mechanisms
apart from applications and apart from how such mechanisms are realized in
the human or in animals.

	The Lighthill report does not recognize this even as a proposition
to be refuted.  Therefore, it is not surprising that Lighthill can't see
the point in experiments designed to study intellectual mechanisms apart
from applications or modelling human or animal behavior.

	In order to refute Lighthill, it should be sufficient to show that
their are real questions about intellectual mechanisms and that
it is worthwhile to perform experiments to answer them.  It would also
help to show that this study has had some success.

	Here are some of the intellectual mechanisms, some questions concerning
them, and some results which have been obtained.

	a. Tree search.

	b. Discrimination.

	c. Representation of information.
		About static scenes
		About laws of motion
		About states of mind

	d. Modelling

	e. The epistemological and heuristic parts of the AI problem.

	f. Induction.

	g. Learning and being taught.


	2. Remarks

		a. Chess is the drosophila of artificial intelligence - M.A. Kronrod

		b. Would Lighthill advocate doing genetics on elephants?

		c. Perhaps we humans should keep the determination of what
problems are worth solving to ourselves for the time being.

		d. That computers play chess only as well as good amateurs
after 25 years is not necessarily a reason for disappointment.

		e. To make machines as intelligent as humans may take a long
time.  Why should you say the field has failed after 25 years?

		f. It may be more fruitful to psychoanalyze the decision to have
a hydrodynamicist analyze AI than to psychoanalyze why people want to program
intelligence.

		g. A decision to abandon work in artificial intelligence would
be a decision that it is impossible or not worthwhile to understand
intellectual mechanisms.

		h. The British AI groups are doing well considering their weak
computer resources.

		i. The biggest weakness in current AI research is the
"look ma, hands" syndrome.  This is the failure to relate experiments to
any kind of theory of intellectual mechanisms.

		j. Lighthill's goof on chess illustrates the need for
experiment, since others made exactly the same goof and were only
convinced by experiment.

		k. Chess programs lose, because they look at the wrong
variations, not because they misevaluate the resulting positions.


	3. Legitimate beefs against AI.

	Does the public have any legitimate complaints about
how the AI community has spent its money?

		a. The experiments have not sufficiently contributed
to making a theory, because most experimenters don't try.  They
Too many experimenters are satisfied to make a computer do
something no machine has done before.  They don't design their
experiments to get information about intellectual mechanisms
except to determine the adequacy of the mechanisms they have
programmed to carry out a particular task.

	British AI is no worse than any other, although perhaps there
has not yet been a British contribution to the theory of AI comparable
to the British contribution to the theory of infinitesimal calculus.