COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	%laudat[s90,jmc]		Notes on the Laudatio for Madrid honorary degree
C00005 ENDMK
C⊗;
%laudat[s90,jmc]		Notes on the Laudatio for Madrid honorary degree

p1
I didn't become interested in AI when I entered Caltech.
I became interested in the subject only after attending the
Hixon Symposium in September 1948.  I didn't take part, since
I didn't speak or even ask a question.  I was just in the audience.

I doubt that the Inversion of functions defined by Turing machines
had much influence.  It clarified a few points, but didn't define
an approach to AI that could be carried out.  The 1959 Programs
with Common Sense was influential.

p9
My original version of alpha-beta was heuristic, because it relied
on optimistic and pessimistic evaluations of positions.  It was
Michael Levin and Timothy Hart who first made alpha-beta directly
comparable with minimax and who proved that it gave the same result.
However, they didn't publish a paper, and the first publication
was in 1963 by Brudno.

p11
As I recall, Levin's result was conclusive.  Knuth's
contribution, as I recall, was to give the behavior of the
algorithm when assumptions are made about the order in which
moves are examined.  Unfortunately, Knuth's assumptions are too
pessimistic, because he assumes the moves will be examined in
random order, whereas game-playing programs have move-ordering
heuristics that try to take more advantage of alpha-beta by
ordering moves in order of quality.  This is somewhat difficult
to treat mathematically.