COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	@make(letterhead,Phone"497-4430",Who"John McCarthy", Logo Old, Department CSD)
C00005 ENDMK
C⊗;
@make(letterhead,Phone"497-4430",Who"John McCarthy", Logo Old, Department CSD)
@style(indent 8)
@blankspace(6 lines)
@begin(address)
Mr. Bruce Murray
215 Wilton Ave.
Palo Alto, CA 94306
@end(address)
@greeting(Dear Mr. Murray:)
@begin(body)
	I don't find your idea that visual perception depends on
physical perception very plausible.  I have a certain visual concept
of the Shuttle Columbia and would recognize it if I saw it.
Playing with a model or touching the real thing wouldn't much
improve my ability to recognize it.  Our notion of some particular
physical object or type of object is more abstract - neither visual
nor tactile.  We are good at correlating them, and can often
recognize objects visually which we have previously encountered
only tactilely and vice versa.  Experiment: Buy a model of the
Shuttle in a toy store and present it to someone who keeps his
eyes closed.  I'll bet he'll recognize it if he is familiar with
pictures of the Shuttle.

	Well, this is just as speculative on my part as you were
in your letter.

	I don't know of efforts to apply tactile notions of
perception in AI.  It might be worthwhile but probably is difficult.
Binford's group has or soon will have hands with enough of a tactile
sense for this and plan to try to program manipulations mediated
entirely by touch.  I suspect, however, that they will find it
preferable to use the same abstract notion of an object that they
use for vision.  Basically, the permanent features of an object
are abstract and determine how the object is perceived either
visually or tactilely in a particular situation.  Neither the
specific visual experience of seeing it at a particular angle
nor the tactile experience of touching particular parts of it
are as fundamental.
@end(body)
Sincerely,




John McCarthy
Professor of Computer Science