perm filename NOM.B[AM,DBL] blob
sn#716727 filedate 1983-06-29 generic text, type C, neo UTF8
COMMENT ā VALID 00003 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 STATEMENT OF RESEARCH INTERESTS
C00006 00003 COMMITMENT TO ACADEMIC CAREER
C00007 ENDMK
Cā;
STATEMENT OF RESEARCH INTERESTS
My main research interest is Discovery: can we understand how people
synthesize new ideas? I test my hypotheses about human creativity by
building computer programs (AM and Eurisko) which attempt to make
discoveries. Sometimes the results are just what we expected (as with
AM's use of analogy), sometimes they are amusing (as when Eurisko
redefined the criteria for interestingness and then declared that what it
had already done was interesting), and occasionally they are gratifying
(as when Eurisko's unconventional fleet design won the 1981 and 1982
Traveller TCS tournaments) or even electrifying (as when AM rediscovered
cardinality, or when Eurisko discovered what we expect to be the
fundamental primitive device in 3D VLSI technology). Experimenting with
these programs leads to criticism and improvement of the original
hypotheses, to a slightly deeper understanding of human creativity, and
thus to the next round of improvements in the programs. The long-range
goal of such research is to exploit the synergy between man and machine,
tapping into the differences in processing abilities and limitations of
each.
Many issues must be dealt with in constructing such programs: how to
represent common sense and expert knowledge concretely; how the program is
to judge the worth of new concepts and conjectures it proposes; how to
constrain the generation of such concepts and conjectures to a small set
of highly plausible and promising ones. The programs must make tactical
decisions about when to reason symbolically and inductively (and slowly)
versus quickly using only statistical methods.
During the 1983-5 period, my specific plans for extending the machine
discovery programs are (i) to recode them in such a fashion that other
researchers in machine learning can easily inspect them and use parts from
them in their own work, (ii) to add a sufficient man-machine interface to
enable researchers outside our narrow subfield to use the program as a
black box "learning module" for their own systems, (iii) continue having
Eurisko perform in difficult task areas, specifically the design of VLSI
circuits, and (iv) use the results of the first three activities to extend
the nascent Theory of Heuristics we developed during the past two years.
COMMITMENT TO ACADEMIC CAREER
After receiving my PhD in 1976, I became an assistant professor at
Carnegie-Mellon University, and then assumed that position at Stanford
University. I have designed and introduced several new courses here and
at CMU, and have enjoyed teaching, advising, and the "idea amplification"
that is only possible within academia. I hope to receive tenure at
Stanford, and have clearly made a commitment to an academic career.