Copy Link
Add to Bookmark
Report
AIList Digest Volume 2 Issue 034
AIList Digest Thursday, 22 Mar 1984 Volume 2 : Issue 34
Today's Topics:
Corporate AI - Entry Route Request,
AI Documents - HAKMEM Request,
Inference - Identifying Programs,
Fuzzy Sets - Reference,
Computer Art - Computer Manipulated Novel,
Expert Systems - Computer EKG's,
AI Funding - Strategic Computing in the New York Review of Books,
Public Service - Tax Info,
Seminars - RUBRIC: Intelligent Information Retrieval &
Computational Linguistics &
Expert System for Building Expert System Rules
Course Announcement - Lisp: Language and Literature
----------------------------------------------------------------------
Date: 19 Mar 84 19:06:44-PST (Mon)
From: decvax!dartvax!lorien @ Ucb-Vax
Subject: IBM vs. HP: research (AI) question
Article-I.D.: dartvax.922
I have been offered entry-level positions at both Hewlett-Packard and
IBM. I feel that, sooner or later, I'd like to do research in some AI-
related field, and I'd like any comments you may have as to the
accessibility of the research labs to an employee starting out as a
programmer in either company. I don't want to start a ridiculous
discussion of the overall merits of and/or problems with HP and IBM;
many articles have been written on both. But things can change quickly
and there may be some work being done of which I'm not presently aware.
I'd appreciate any impressions, subjective or otherwise, that you may
have. I hold an A.B. in Computer Science from Dartmouth.
--Lorien Y. Pratt
Dartmouth College Library
Hanover, NH 03755
decvax!dartvax!lorien
------------------------------
Date: 10 Mar 84 18:13:23-PST (Sat)
From: pur-ee!ecn-ee!davy @ Ucb-Vax
Subject: Copies of HAKMEM? - (nf)
Article-I.D.: pur-ee.1672
This has already been asked in UNIX-WIZARDS, I thought I'd ask it here
too. Does anyone have a copy of HAKMEM (MIT Memo from Feb. 1972) they'd
be willing to Xerox? I've heard there's an online copy at MIT-MC or
someplace -- anyone know where it's at?
--Dave Curry
decvax!pur-ee!davy
eevax.davy@purdue
[I have a copy of this memo (AIM 239, HAKMEM by M. Beeler, R.W. Gosper,
and R. Schroeppel). It is a collection of notes by MIT hackers on about
20 different topics. The document is about 100 pages long and includes
figures. Does an online copy exist? -- KIL]
------------------------------
Date: Mon 19 Mar 84 23:50:08-PST
From: PEREIRA@SRI-AI.ARPA
Subject: Identifying programs
"Algorithmic Program Debugging" by Ehud Shapiro, MIT Press includes
substantial discussion of the question of identifying programs
from I/O pairs. Of course in general the identification is not
exact. Concepts of asymptotic identification ("identification in
the limit") are used instead. A lot of this work has been
developed to try to pin down the concept of "learnable language".
There are a number of recent papers on this question by Scott
Weinstein (University of Pennsylvania) and others, in the journal
Information and Control. If anyone is interested, I'll dig out
the references.
-- Fernando Pereira
------------------------------
Date: 19 Mar 1984 20:56:34-PST
From: don%brandeis.csnet@csnet-relay.arpa
Subject: Fuzzy Sets
I have a reference for Daniel Conde who requested information about
Fuzzy Sets on a recent AI bulletin board:
Fuzzy Sets and Systems: Theory and Applications
Didier Dubois & Henri Prade
copywrite 1980
Academic Press
Don Ferguson
------------------------------
Date: 19 Mar 84 11:36:10 CST (Mon)
From: ihnp4!houxa!homxa!rem
Subject: Computer Manipulated Novel
For people interested in computer-aided art, I manipulated a small,
unpublished novel of mine called ABRACADABRA a few years ago. The
book is a mystery derived from childhood experiences in St. Louis.
I call the manipulated book ABRACADABRA CADAVER. Chapter-by-chapter
I wrote UNIX shell programs to alter the text according to its con-
tents: for example, in an early chapter I misspelled all words as
a child might do. In another I inserted German proverbs appropriate
to my father's speech in all of his conversations. Another repeats
key phrases again and again, in a minimalist way; another puts all
dialog into footnotes; another, where the mystery unfolds, cryptically
reverses the sentences throughout--and so on. After editing the
end results, I came up with a Joycean-like book that is quite
readable and interesting as a literary document. I no longer
have it on-line, but if anyone is interested, I can provide
more details. And, of course, if anyone knows of a publisher
crazy enough.....
Bob Mueller
BELLCORE
Holmdel, NJ
------------------------------
Date: 11 Mar 84 11:01:10-PST (Sun)
From: harpo!ulysses!burl!clyde!akgua!mcnc!ecsvax!hsplab @ Ucb-Vax
Subject: Computer EKG's
Article-I.D.: ecsvax.2145
One reason why computerized EKG's have become so popular in the medical
environment is that **most** of the EKGs are performed on normal people
and are being used as a screening process. This means that if a computer
program is very good at differentiating between normals and abnormals
without any other capability (not true with current programs), it will
probably do better than 90%. It is for this reason that a cardiologist
overview is used primarily to catch gross errors and to refine problems
associated with pathological cases. In a study done by Bailey at the
NIH in the early 1970's, most computer programs actually did rather well,
and if you removed interpretation differences which were common among
cardiologists and tested the programs on grossly abnormal cases, they
were able to achieve better than 60%-70% accuracy.
David Chou
Department of Pathology
University of NC, Chapel Hill
!decvax!mcnc!ecsvax!hsplab
------------------------------
Date: 20 Mar 84 11:46:20 PST (Tuesday)
From: Ron Newman <Newman.es@PARC-MAXC.ARPA>
Subject: Strategic Computing in the New York Review of Books
The March 15, 1984 issue of The New York Review of Books contains an
article entitled "The Costs of Reaganism", which mentions DARPA's
Strategic Computing Program as an example of misdirected U.S. economic
and budgetary policy. The article is by Emma Rothschild, who teaches in
the Science, Technology, and Society program at MIT and is the author of
"Paradise Lost: The Decline of the Auto-Industrial Age".
...What does it mean for America's future economic growth that
69 percent of federally supported research and development is
for military purposes, an increase since 1981 of $18.1 billion
in the military function and of $0.6 billion in non-military
functions? [21]
Does it matter for the character of America's scientific
institutions that the Defense Advanced Research Projects
Agency's new "strategic computing" program is in the process
of transforming academic computer science?[22] Does it
matter for American competitiveness that Japan's ten-year
program on the cognitive, linguistic, and engineering
foundations of computing will be civilian, while America's
will be concerned with robot reconnaissance vehicles,
radiation-resistant wafers, and missile defenses, with
"speech recognition" in the "high-noise, high-stress environment
[of] the fighter cockpit," and with "voice distortions due
to the helmet and face mask"? [23] Mr. Reagan's principal
opponents are not asking these questions; they are questions
about the militarization of the political life, the scientific
potential, and the economic society of the richest country in
the world.
[21] "Special Analyses", Budget of the United States Government,
FY 1985, p. K-30.
[22] The program is described in Weinberger's Annual Report, p. 263,
and also in the Defense Advanced Research Projects Agency's
own study "Strategic Computing" (DARPA, October 28, 1983).
In this study DARPA explains that it intends to use contract
personnel from industry as well as university researchers, in
order to "avoid a dangerous depletion of the university
computer science community": "The magnitude of this national
effort could represent a very large perturbation to the
university community" (p. 64)
[23] DARPA, "Strategic Computing", pp. 34-35.
------------------------------
Date: Wed 21 Mar 84 14:05:12-PST
From: Ken Laws <Laws@SRI-AI.ARPA>
Subject: Tax-Free Support vs. Income Averaging
Bob Boyer of UTexas-20 posted a bboard message about IRS policy on
tax-free student fellowships. This isn't AIList material, but it
will be of interest to many students on and off the Arpanet, so I
am making it available for those who want to post it at their sites.
I have copied the message to file <AIList>IRS.TXT on SRI-AI, and will
send copies to interested people who can't FTP the file. I have
also included related messages from others who read the original.
The content is roughly this: if you claim that your current academic
support is tax-free (or if the IRS makes that claim), and if such income
is at least 50% of your support, you will probably not be able to income
average during the next four years. This is very likely to cost you more
than the tax you save on your current fellowship or other support.
-- Ken Laws
------------------------------
Date: 19 Mar 84 14:23:22 PST (Monday)
From: Kluger.PA@PARC-MAXC.ARPA
Reply-to: Kluger.PA@PARC-MAXC.ARPA
Subject: HP Computer Colloquium 3/22
[Forwarded from the SRI-AI bboard by Laws@SRI-AI.]
Richard M. Tong
Advanced Information and Decision Systems
RUBRIC: An Intelligent Aid for Information Retrieval
In this talk I will describe an ongoing research project that is
concerned with developing a computer based aid for information retrieval
from natural language databases. Unlike other attempts to improve upon
Boolean keyword retrieval systems, this research concentrates on
providing an easily used rule-based language for expressing retrieval
concepts. This language draws upon work in production rule systems in AI
and allows the user to construct queries that give better precision and
recall than more traditional forms.
The talk will include a discussion of the main elements in the system
(which is written in LISP and C), the key research issues (including
some comments on the important role that uncertainty plays) and some
man-machine interface questions (in particular, the problem of providing
knowledge elicitation tools).
Thursday, March 22, 1984 4:00 pm
*** Please note the location change ***
Hewlett-Packard
1651 Page Mill Road
Palo Alto, CA
28C Lower Auditorium
Be sure to arrive at the building's lobby on time, so that you may be
escorted to the meeting room
------------------------------
Date: 19 Mar 1984 15:26 EST (Mon)
From: "Daniel S. Weld" <WELD%MIT-OZ@MIT-MC.ARPA>
Subject: Computational Linguistics (BOSTON)
[Forwarded from the MIT bboard by SASW@MIT-MC.]
Wednesday, March 21 4:00pm 8th floor playroom
De-mystifying Modern Grammatical Theory and Artificial Intelligence
Robert Berwick
It has frequently been suggested that modern linguistic theory is
irreconcilably at odds with a ``computational'' view of human
linguistic abilities. In fact, linguistic theory provides a rich
source of constraints for the computationalist. In this talk I will
outline some of the key changes in grammatical theory from the mid 60's to
the present day that support this claim, and at the same time try to
dispel a number of myths:
Myth: Modern grammars are made up of large numbers of rules that
one cannot ``implement.''
Myth: Modern grammars are not relevant to computational models
of language processing.
Myth: Knowledge that you can order hamburgers in restaurants
aids *on-line* syntactic processing.
------------------------------
Date: 20 Mar 84 11:30:18 EST
From: Smadar <KEDAR-CABELLI@RUTGERS.ARPA>
Subject: Experiments with Rule Writer for EXPERT
[Forwarded from the Rutgers bboard by Laws@SRI-AI.]
I I I SEMINAR
Title: Experiments with Rule Writer for EXPERT
Speaker: George Drastal
Date: Tuesday, April 3, 1984, 1:30-2:30 PM
Location: Hill Center, Seventh floor lounge
George Drastal, a Ph.D. student in our department, will describe his thesis
reseach in an informal talk. His abstract:
Results are presented of some experiments with Rule Writer, an AI
system that assists knowledge engineers with the task of writing
inference rules for a medical consultation system in the EXPERT
formalism. Rule Writer (RW) is used primarily in an early stage of
expert system development, to generate a prototype rule base. RW may
also be used as a testbed for experimenting with alternative
organizations of expert knowledge in the EXPERT knowledge
representation.
------------------------------
Date: Wed, 21 Mar 84 18:38 PST
From: BrianSmith.PA@PARC-GW.ARPA
Reply-to: BrianSmith.PA@PARC-GW.ARPA
Subject: Course Announcement -- Lisp: Language and Literature
[Forwarded from the SRI CLSI bboard by Laws@SRI-AI.]
The following course will be the CSLI Seminar on Computer Languages
for the Spring Quarter [at Stanford]. If you are interested in attending,
please read the notes on dates and registration, at the end.
Lisp: Language and Literature
A systematic introduction to the concepts and practices of programming,
based on a simple reconstructed dialect of LISP. The aim is both to
convey and to make explicit the programming knowledge that is
typically acquired through apprenticeship and practice. The material
will be presented under a linguistic reconstruction, using vocabulary
that should be of use in studying any linguistic system. Considerable
hands-on programming experience will be provided.
Although intended primarily for linguists, philosophers, and
mathematicians, anyone interested in computation is welcome. In
particular, no previous exposure to computation will be assumed.
However, since we will aim for rigorous analyses, some prior familiarity
with formal systems is essential. Also, the course will be more like a
course in literature and creative writing, than like a course in, say,
French as a second language. The use of LISP, in other words, will
be primarily as a vehicle for larger issues, not so much an object of
study in and of itself. Since LISP (unlike French) is really very
simple, we will be able to teach it in class and lab sessions. Tutorial
instruction and some individual programming assistance will be provided.
Topics to be covered include:
-- Procedural and data abstraction;
-- Objects, modularity, state, and encapsulation;
-- Input/output, notation, and communication protocols;
-- Meta-linguistic abstraction, and problems of intensional grain;
-- Architecture, implementation, and abstract machines;
-- Introspection, self-reference, meta-circular interpreters, and reflection.
Throughout, we will pay particular attention to the following themes:
-- Procedural and declarative notions of semantics;
-- Interpretation, compilation, and other models of processing;
-- Implicit vs. explicit representation of information;
-- Contextual relativity, scoping mechanisms, and locality;
-- Varieties of language: internal, external, theoretical;
-- Syntax and abstract structure: functionalism & representationalism.
Organizational Details:
Instructor: Brian C. Smith, Xerox PARC/Stanford CSLI; 494-4336 (Xerox);
497-1710 (Stanford), "BrianSmith@PARC" (Arpanet).
Classes: Tuesdays and Thursdays, 2:00 - 3:30, in Room G19, Redwood
Hall, Jordan Quad.
NB: Since we will be using the computers just now being installed
at CSLI, there may be some delay in getting the course underway.
In particular, it is possible that we will not be able to start until
mid-April. A follow-up note with more details will be sent out as
soon as plans are definite.
Registration: Again, because of the limited number of machines, we
may have to restrict participation somewhat. We would therefore
like anyone who intends to take this course to notify Brian Smith
as soon as possible. Note that the course will be quite demanding:
10 to 20 hours per week will probably be required, depending on
background.
Sections: As well as classes, there will be section/discussion periods
on a regular basis, at times to be arranged at the beginning of the
course.
Reading: The course will be roughly based on the "Structure and
Interpretation of Computer Programs" textbook by Abelson and
Sussman that has been used at M.I.T., although the linguistic
orientation will affect our dialects and terminology.
Laboratory: Xerox 1108s (Dandelions) will be provided by CSLI, to be
used for problem sets and programming assignments. Instructors &
teaching assistants will be available for assistance at pre-arranged
times.
Credit: The course may be listed as a special topics course in Computer
Science. However (in case that does not work out) anyone wishing
to take it for credit should get in touch, so that we can arrange
reading course credit.
------------------------------
End of AIList Digest
********************
