Copy Link
Add to Bookmark
Report
AIList Digest Volume 3 Issue 127
AIList Digest Monday, 23 Sep 1985 Volume 3 : Issue 127
Today's Topics:
Seminars - Spencer-Brown Seminar in San Francisco &
Knowledge Representation (SRI) &
Knowledge-Directed Database Management (IBM-SJ) &
Learning Spatial Concepts (CSLI) &
2nd-Order Lambda Calculus (UPenn) &
Computer Tutor for Programming Recursion (UCB),
Course - AI Theories of Belief and Action (SU)
----------------------------------------------------------------------
Date: Wed, 18 Sep 85 13:15:23 pdt
From: william@aids-unix.ARPA (william bricken)
Subject: Spencer-Brown seminar in San Francisco
G. Spencer-Brown, the author of LAWS OF FORM, will be presenting
a five-day seminar at the Miyako Hotel in San Francisco, from
October 7 through 11, 1985.
Morning sessions will address the technical methodology of Laws
of Form. Afternoon sessions will address "open intelligence",
Spencer Brown's ideas on life and living.
Registration costs: $650 for both sessions,
$300 for afternoons only.
Information: UNI-OPS
260 Marshall Drive
Walnut Creek, CA 94598-2833
phone Walter Zintz, (415) 945-0048
PERSONAL COMMENTS: Spencer-Brown's work on the 4-color map theorem was
discussed in this list early last year. The man is an iconoclast, and
has managed to alienate a significant portion of most audiences. The
afternoon session may be anything from EST training (which Spencer-Brown
is said to have originated) to metaphysical ecstacy. The application of
Brownian mathematics to representation theory and automated theorem
proving in AI is profound. Followers of his work range from mystics
to mathematicians; he is both. I have heard that he is not talking
to mathematicians.
william
------------------------------
Date: Wed 18 Sep 85 15:19:44-PDT
From: LANSKY@SRI-AI.ARPA
Subject: Seminar - Knowledge Representation (SRI)
THE SKY IS A BLUE COLOR
Marcel Schoppers
SRI International AI Center
11:00 AM, MONDAY, September 23
SRI International, Building E, Room EJ228 (new conference room)
I present a representation which will allow us to encode and
access much of the information contained in simple descriptive
statements. "The sky is a blue color" entails that blue is a
color, that the sky has a color, that the sky is blue, that the
sky is visible, and that the sky is located both spatially and
temporally. These generalizations are so trivial that they border
on presuppositions, and they have consequently been taken for
granted in semantic nets and frames. Making such information
explicit greatly increases the density and usefulness of stored
knowledge. One interesting application is to disambiguate an
adjective/predicate to suit a given noun/extension.
My representation is parsimonious, having O(three) primitive con-
structs (link types); is highly irredundant, since blue(sky) and
color(blue) reference the same blue; and is static, being inten-
ded to formalize and implement "massively parallel" deterministic
connectionist question-answering systems. Predicates, relations,
and simple forms of quantification all emerge as by-products of
function applicability and set inclusion. Viewed as a logic the
representation is potentially O(w), intensional and inconsistent.
The talk will touch on issues in philosophy of logic and linguistics.
I will especially appreciate constructive criticism in those areas,
as I am a novice there.
------------------------------
Date: Wed, 18 Sep 85 15:34:00 PDT
From: IBM-SJ Calendar <CALENDAR%ibm-sj.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - Knowledge-Directed Database Management (IBM-SJ)
IBM San Jose Research Lab
5600 Cottle Road
San Jose, CA 95193
CALENDAR
SEPTEMBER 23 - 27, 1985
Thurs., Sept. 26 Computer Science Seminar
10:00 A.M. PROBE: A RESEARCH PROJECT
Aud. B IN KNOWLEDGE-DIRECTED DATABASE MANAGEMENT
Conventional record-based DBMSs will be
inadequate for many of the knowledge-intensive
information processing applications (e.g.,
business and industrial automation, CAD/CAM, and
military command and control) of the future.
These applications require integrated access to
a variety of information types (e.g., images,
maps, signals, text) not currently supported by
DBMSs. Also, they rely on specialized knowledge
or expertise for processing the new information
types; for many of these types, specialized
storage devices and processors (e.g.,
workstations, image enhancers, solid modellers),
are or will be available. Currently, DBMSs have
no general facilities for efficiently
assimilating and utilizing this special
knowledge or for incorporating these specialized
processors into their own processing. The
objective of the PROBE project is to develop an
advanced DBMS effective for these
knowledge-intensive applications. Our approach
is to enhance existing DBMSs with (a)
user-defined object classes as the basis for
defining new information types and operations
and for integrating specialized processors, (b)
dimensional (space and time) concepts, which are
a common characteristic of many of the new
information types, and (c) recursive predicates
and queries, which provide intensional knowledge
processing capabilities essential for many of
the applications. In each case, it is necessary
to augment both the logical (data model, query
language) components and the physical (storage
structures, access methods, query processor)
components of the DBMS. In this talk, we
describe approaches to addressing all these
issues.
Dr. U. Dayal, Computer Corporation of America
Host: C. Mohan
------------------------------
Date: Thu 19 Sep 85 08:14:26-PDT
From: Emma Pease <Emma@SU-CSLI.ARPA>
Subject: Seminar - Learning Spatial Concepts (CSLI)
[Excerpted from the CSLI Newsletter by Laws@SRI-AI.]
``Crossing the Rubicon: From a Physics of Dead Coordinate Spaces
to a Physics of Living Coordinate Spaces''
Dr. Peter Kugler, The Crump Institute for Medical Engineering, UCLA
Monday, September 23, 1985, 2:15pm, Ventura Hall
This talk will be about self-organizing systems that involve
low-energy (nonforce) coupling and the nature of the predicates that
constitute the low-energy descriptors, and will be organized around
issues pertaining to general problems of language and information.
The emphasis will be on systems that generate (self-assemble) new
levels of description. These new levels constitute new languages
parasitic on the lower level languages but not reducible to their
predicates. In the self-organizing systems of interest it is the
``coordinate spaces,'' which are themselves evolving, that become the
important objects of study. Instead of assuming a fixed coordinate
space, when the interest focuses on trajectories, attention is devoted
to the coordinate space itself, since this is what provides the semantics.
This approach is very similar to developments in computer
architecture that focus on parallel processing. In these machines
(connection machines, Boltzmann, etc.) the machine language
self-organizes (e.g. programs itself through the emergence of new
stable configurations), and the new predicate descriptions play the
role of symbols in terms of their opacity with respect to the lower
level language. The machine language `gives birth' to the symbolic
level of description. This situation contrasts dramatically with that
of von Neumann machines, for which the symbolic language is
ontologically independent of the machine language. A symbolic
language can run on any of an infinite variety of mechanistic
substrates, the primacy of the symbol prevailing over the substrate
machine. The approach advocated here, puts the focus on the machine
level of interaction, thus preserving an ontological continuity and
avoiding mind/body, syntactic/semantic, etc. problems.
------------------------------
Date: Fri, 20 Sep 85 11:06 EDT
From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - 2nd-Order Lambda Calculus (UPenn)
THE 2ND ORDER LAMBDA CALCULUS - Dale Miller, Penn CIS
Joint Mathematics / Computer Science Logic Colloquium
4:40 Monday 23 September 1985, DRL 4E17, University Of Pennsylvania
In this talk we will present the description of an extended type lambda
calculus where types can be used as values. We will illustrate how this
language can be used to perform computation in novel fashions and what
semantics might work with it.
------------------------------
Date: Wed, 18 Sep 85 11:12:07 PDT
From: davies%ucbcogsci@Berkeley (Catherine Davies)
Subject: Seminar - Computer Tutor for Programming Recursion (UCB)
BERKELEY COGNITIVE SCIENCE PROGRAM
Fall 1985
Cognitive Science Seminar -- IDS 237A
TIME: Tuesday, September 24, 11:00 - 12:30
PLACE: 240 Bechtel Engineering Center
(followed by)
DISCUSSION: 12:30 - 1:30 in 200 Building T-4
SPEAKER: Peter Pirolli, School of Education,
UC Berkeley
TITLE: ``A Cognitive Model and Intelligent Computer
Tutor for Programming Recursion''
Recursion is typically a novel concept for programming stu-
dents that causes them considerable grief and difficulty.
Thus, the study of how people learn to program recursive pro-
grams provides a useful domain for addressing the psychologi-
cal issue of how fundamentally new knowledge is acquired as
well as the instructional issue of how to teach a difficult
programming concept. I will present a production system model
that addresses expert and novice problem-solving, problem-
solving by analogy, and skill acquisition in programming
recursive functions. This research served as the basis for
the development of recursion lessons in an intelligent com-
puter tutor for programming LISP. Specifically, a simulation
model of ``ideal'' and ``buggy'' novice problem-solving was con-
structed for coding recursion. Using this model, the LISP
tutor provides instruction, hints, and feedback in the context
of programming. The LISP tutor also maintains a model of the
skill development of individual students. Evaluations show
that the LISP tutor is more effective in teaching introductory
LISP programming than good classroom instruction and
approaches the effectiveness of human tutors.
------------------------------
Date: Wed 18 Sep 85 09:26:04-PDT
From: Kurt Konolige <KONOLIGE@SU-SCORE.ARPA>
Subject: Course - AI Theories of Belief and Action (SU)
[Forwarded from the Stanford bboard by Laws@SRI-AI.]
Announcement of a new course in Artificial Intelligence:
CS429: Formal AI Theories Of Belief and Action
Instructor: Kurt Konolige, SRI International and CSLI
Description: We will discuss some formal commonsense theories of
belief and action that have emerged over the last 5-7 years. This
course is intended both as a survey of recent research in this area,
and as an introduction to the application of techniques from formal
logic to AI problems. The emphasis is on acquiring a facility for
working with the tools of logic, especially by analyzing and
critiquing current AI research. The following is a list of topics we
will cover:
1. Knowledge and Belief
- Kripke and sentential semantics
- Modal logics for knowledge and belief
- Proof methods for quantified modal logics,
including recent resolution methods
2. Formal planning theories
- The situation calculus
- Integration of action and belief
3. Further topics
- Introspective belief theories
- The relation between introspection and nonmonotonic
reasoning
Prerequisites: Some familiarity with basic concepts from first-order
logic will be assumed, e.g., students should know what what a model
is, how to construct proofs from axioms and rules of inference, and so
on.
Requirements: Graduate-level course for CS students, 3 units.
Students will be expected to read 1-2 papers a week. There will be
one problem set a week, and either a project or final exam.
Time: MWF 11am (50 min), MJH 352, Fall 1985
------------------------------
End of AIList Digest
********************