COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	INITIAL STATEMENT
C00005 ENDMK
C⊗;
INITIAL STATEMENT


	Professor Lighthill finds pointless and unsuccessful a class
of research he calls building robots.  By a robot he means a device
or computer program designed to "mimic a human function" without being
useful or without being an attempt to simulate the way in which
humans perform the function.

	I think his objection is related to the mistaken way he classifies
artificial intelligence into three categories.  The mistake is that he
report doesn't recognize that people are trying to study intellectual
mechanisms apart from applications and apart from attempts to determine
how humans or animals carry out intellectual functions.  I say the report
doesn't recognize this, because it doesn't mention the idea even to
disagree with it, although the idea is explicit in the works of many
of the authors to whom he refers, and in the letter from Professor Meltzer
that he explicitly solicited.
I fear that Lighthill's complaints about the bad style of the literature
in artificial intelligence is his way of confessing that he hasn't read
much of the material he solicited.

	In order to show that Lighthill is mistaken, it seems sufficient
to show that there are interesting questions about intellectual processes
that can usefully be studied apart from applications, and that experiments
are useful in this study.

	In order to do this, I shall try to condense two quarters of
lectures on artificial intelligence into ten minutes without talking fast.
I shall mention some intellectual processes that have been studied and
some of the experiments that have been made.  Besides this, I shall mention
some difficult problems that remain.  The examples chosen are ones that
can be described briefly.