perm filename KNUTH[AM,DBL] blob
sn#244077 filedate 1976-10-22 generic text, type T, neo UTF8
Here is the text of a letter now winging its way to you via the mails:
Dear Don,
Greetings from the East. Merle and I have adjusted well to life here.
After bracing for a cultural shock, expecting the worst, we've been
very pleasantly surprised over and over again at how livable
Pittsburgh is. (e.g., at the zoo, one can watch a flock of penguins
swimming about: from the side, they appear to be normal birds flying
in slow motion through the air. As another example, the symphony
turned out to be a top-rated one). The only danger appears to be the
very mixed quality of restaurants around here. The CMU CS facilities
are enough like home for me not to grumble too much about adjusting,
and the people in the department are super.
Mike Shamos and I have discussed possible applications of AI
techniques to computational geometry. That appears to be non a
propos , however, because that field is too knowledge-poor. For
example, the concepts it uses (e.g., "spanning lines") were those
developed long ago, for very different pursuits. There may actually
be room for a concept-isolater like AM to move in and define some
new, potentially useful concepts. Do you have any feelings about the
liklihood of that project getting anywhere?
This year I'm beginning a large but loosely coupled collection of
research efforts, programs which attack various aspects of mechanical
mathematics. Included is work on a theorem-prover which learns (by
teaching, discovery, and analogy), a new improved version of AM, a
Meta-AM system which peers over AM's shoulder and occasionally
suggests new heuristics for it to use, an analogizer, and a librarian
who carefully archives the past experiences of each module in the
combined system. Naturally, my role on most of these projects is as
adviser or co-researcher only. Herb Simon is working closely with me
on the codification of meta-rules (heuristics for filling in new
heuristics) -- we've already been surprised by the problem: it
appears to be tractablle.
Also, I'm splitting the thesis up into a few papers, which of course
absorbs an uncomfortable amount of time and energy.
I've only had to give a couple lectures this semester; they don't put
new faculty to work until second semester. Even then, all I'll be
responsible for is the AI Seminar.
All of us in the department are very anxious to have you come out for
a visit sometime. You can have carte blanche , in all aspects of
your trip: who comes with you, where you stay, how much
privacy/interaction you have, how ong you stay, etc. I'd personally
love to be able to talk over my plans for the next year's research
with you, and Shamos, Kung, Traub would, too. Bentley will be coming
to CMU starting next semester. Also, don't forget our math dept.,
which is oriented along algorithmic lines (Fix, Andrews, etc.). I
hope you'll decide to come out for a week or two -- or more -- this
year.
Finally, let me mention that our department has the means to hire
roughly one new faculty member each year. If any of your students are
graduating, consider reccommending them to us (and vice versa).
Regards,
Doug Lenat