perm filename LET.E3[AM,DBL] blob
sn#373547 filedate 1978-08-16 generic text, type T, neo UTF8
To: Dr. Eamon Barrett, NSF
From: Dr. Douglas B. Lenat, Carnegie-Mellon University
Date: July 7, 1978
Subject: Update to Research Proposal
The research I intend to pursue during the coming year (July 1, 1978
through June 30, 1979) is essentially unchanged from that outlined in
my previous proposal [10] (MCS-7704440: "The use of informal rules to
guide the search for discoveries in Mathematics".). How can we
account for the creative syntheses of new mathematical concepts, for
the facility which human mathematicians demonstrate at finding
proofs, deciding quickly the promise of various new definitions,
noticing conjectures, etc.?
This research followed closely upon the construction of a computer
program (AM) which carried on simple mathematical theory formation on
its own. This is summarized in the proposal [10], and discussed at
varying lengths in [1]-[9].
In my two-year proposal, I intended to (i) get a handle on the
problem, (ii) build a prototypical system to use an experimental
vehicle, (iii) re-design from that experience, tentatively along the
lines of change of representation. By now we have accomplished (i)
and are well along on (ii), although the "handle" we discovered is
quite unlike what was expected (as discussed in the annual report on
this project): We represent each heuristic rule as a full-fledged
concept with facets (a frame with slots, a property list with
properties), and treat heuristics as if they were a special sort of
mathematical operation, nothing more.
I am leaving Carnegie-Mellon University to become an Asst. Professor
at Stanford University. Effective August 1, my research will be
conducted from there. The primary function of this document is to
discuss the changes -- and absence of changes -- necessitated by such
a transition. I shall continue my current line of investigation --
into the mechanisms whereby heuristic rules can be synthesized and
modified -- and I anticipate no interruption in that endeavor.
Activities (ii) and (iii) in the paragraph above will be continued.
The facilities at Stanford are, for my purposes, equal to or surpass
those at CMU. For example, Interlisp is present, as is a
two-dimensional screen editor, Xerographic printers, etc. On a more
personal level, the valuable interactions I am giving up (with
Herbert Simon) will be replaced by other regular research discussion
meetings (with Ed Feigenbaum, Donald Knuth, Bruce Buchanan, and other
members of the Stanford Computer Science and Mathematics
Departments). The graduate student assistance I require can and will
be met by Stanford graduate student research assistants. The
equipment I require is being purchased by CMU from this grant, and
those funds will be used by Stanford to re-purchase equivalent
equipment for this project.
The paramount consideration effecting my decision to shift my
appointment to Stanford was the research environment it provided:
the increased expected yield of my research. The shift will cause a
delay in the timetable, but not one of more than a few weeks. This
lack of trauma is due to the fact that I spent the years 1972-6 as a
graduate student at Stanford, and am familiar with the systems,
languages, staff, faculty, and many of the graduate students there.
I am receiving a great deal of support from the department chariman
(Professor E. A. Feigenbaum), department administrators (Ms. B.
Scott), and others, to ease my transition as much as possible.
Published papers about the AM and Eurisko research projects
[1] AM: An Artificial Intelligence Approach to Discovery in
Mathematics as Heuristic Search, Ph.D. Thesis, Stanford A. I. Lab
Memo AIM-286, CS Report No. STAN-CS-76-570, and Heuristic
Programming Project Report HPP-76-8, Stanford University, July, 1976.
[2] Designing a Rule System That Searches for Scientific Discoveries,
(Lenat and Harris), invited paper for the PDIS conference in
Honolulu, May, 1977; published in (Hayes-Roth and Waterman, eds.)
Pattern-Directed Inference Systems, Academic Press, 1977. Also
issued as a CMU technical report, April, 1977.
[3] Automated Theory Formation in Mathematics, Fifth IJCAI,
Cambridge, Mass., August, 1977.
[4] Less Than General Production System Architectures, (Lenat and J.
McDermott,) Fifth IJCAI, Cambridge, Mass., August, 1977.
[5] The Ubiquity of Discovery, the 1977 Computers and Thought Lecture
(invited talk at the Fifth IJCAI). Preliminary version published in
the proceedings of that conference; final version printed in the
Journal of A.I. Reprinted in New York's The Sciences, and repeated
as an invited talk at NCC (Anaheim, June, 1978).
[6] On Automated Scientific Theory Formation: A Case Study Using the
AM Program, invited paper presented at the Ninth Machine Intelligence
workshop in Leningrad, USSR, April, 1977. Forthcoming publication in
(Michie, ed.) Machine Intelligence 9, 1978.
[7] Artificial Intelligence and Natural Statistics, invited paper at
"Computer Science and Statistics: Eleventh Annual Symposium on the
Interface", University of North Carolina at Raleigh, March 6, 1978.
[8] Unscripted interview on AI & Problem Solving, broadcast over the
BBC, as part of the Open University's 32 week course on Cognitive
Psychology. Taped at CMU on Feb. 22, 1978, by Clive Holloway, Open
University, Milton Keynes, England.
[9] Expert Systems, a section of the forthcoming COSERS report, 1978.
[10] The Use of Informal Rules to Guide the Search for Discoveries
in Mathematics, Research Proposal MCS-7704440, submitted to NSF,
November 9, 1976.