AIList-REQUEST@SRI-AI.ARPA (AIList Moderator Kenneth Laws) (11/22/85)
AIList Digest Friday, 22 Nov 1985 Volume 3 : Issue 175
Today's Topics:
Seminars - Unification Revisited (SRI) &
Expanding the Horizons of Expert Systems (SRI) &
CYC Commonsense Knowledge Project (GTE) &
A Multimodal Perceptual System (UPenn) &
ANALOGICA '85 (Rutgers),
Conference - Foundations of AI
----------------------------------------------------------------------
Date: Tue 19 Nov 85 10:02:51-PST
From: OLENDER@SRI-AI.ARPA
Subject: Seminar - Unification Revisited (SRI)
DATE: Monday, November 25, 1985
LOCATION: EJ242
UNIFICATION REVISITED
Jean-Louis Lassez
IBM Thomas J. Watson Research Center
There are three main approaches to finitely represent sets of
solutions of equations in the Herbrand Universe. In Robinson's
classical approach the set of solutions is represented by an mgu which
is computed from the set of equations. We introduce a dual approach,
based on Plotkin's and Reynold's concept of anti-unification in which
the finite representation (mgs) is now "lifted" from the set of
solutions. A third approach proposed by Colmerauer is based on the
concept of eliminable variables.
The relationships between these three approaches are established.
This study provides an appropriate setting to address the problem of
solving systems of equations and inequations which arises in recent
extensions to Prolog. A key result is that the meta-equation
E = E1 v E2 v ... v En
admits solutions only in trivial cases. Two important corollaries
follow naturally. The first is Colmerauer's property of independences
of inequations. This means that deciding whether a system of
equations and inequations has solutions can be done in parallel. The
other corollary is a negative result; the set of solutions of a system
of equations and inequations can be finitely represented by mgu's only
in trivial cases. Consequently, one cannot obtain a simplified system
which is in "solved" form. This is unlike the case when only
equations are considered. Similar properties hold in inductive
inference when one attempts to generalize from sets of examples and
counter-examples.
------------------------------
Date: Tue 19 Nov 85 15:25:48-PST
From: ICHIKI@SRI-AI.ARPA
Subject: Seminar - Expanding the Horizons of Expert Systems (SRI)
EXPANDING THE HORIZONS OF EXPERT SYSTEMS
Piero P. Bonissone and Allen L. Brown, Jr.
General Electric Corp. R&D
Conference Room EK242
Thursday, November 21, 1985 (10:00 - 11:00a.m.)
Abstract: In this paper we analyze the complexity of problem domains,
such as maintenance problems, typically handled by first generation
expert systems. DELTA/CATS-1, an expert system for troubleshooting
diesel electric locomotives, is described as a typical example of such
systems. More complex domains of expertise involving time-varying,
partial and uncertain information cannot be addressed by the
techniques common to first generation expert systems. LOTTA, a
symbolic simulator of battlefield situations, illustrates some of the
requirements of such an intricate domain. We discuss the new
inference techniques required to address the problem of managing
battlefield strategies and tactics. These techniques include
capabilities for reasoning with uncertain and incomplete information.
The state of the art and current research thrusts are discussed.
------------------------------
Date: Thu, 21 Nov 85 11:16:19 EST
From: Bernard Silver <SILVER@MIT-MC.ARPA>
Subject: Seminar - CYC Commonsense Knowledge Project (GTE)
GTE Laboratories AI Seminar
Monday, November 25, 3 pm
Room 8-2335
GTE Laboratories
40 Sylvan Rd.,Waltham, MA
Mayank Prakash of MCC, Austin, Texas, will discuss the current state of
research on the CYC Project.
The CYC Project
The CYC Project at MCC is an attempt to enhance the power of AI
systems by providing them with common sense knowledge. Our approach
is based on the observations that 1) the brittleness of current
AI systems is due to a lack of knowledge of the broader context of
their narrow domain, and 2) automated acquisition of knowledge
requires the system to start out with a critical mass of general
knowledge. Our goal is to create a system with a large common
sense knowledge base. It will serve as a source of power for other
AI systems by acting as a knowedge server and thus providing a deeper
understanding of the domain, and by providing the ability to
analogise. It is our belief that this will significantly enhance the
performance of such systems.
For more information, please contact Bernard Silver (617) 466-2663
------------------------------
Date: Thu, 21 Nov 85 12:11 EST
From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - A Multimodal Perceptual System (UPenn)
Forwarded From: Sharon Stansfield <Stansfield@UPenn>
Date: Thu, 21 Nov 85 11:09 EST
THESIS PROPOSAL:
A RUDIMENTARY ACTIVE,
MULTIMODAL, INTELLIGENT SYSTEM
FOR OBJECT CATEGORIZATION
S. A. Stansfield
This proposal outlines the design of a knowledge-based, active, multimodal
perceptual system for the task of object categorization. Passive stereo vision
and active touch are used.
Basic level, or generic, objects will be recognized. The representation for
reasoning will be hierarcical and frame-based. At the highest level will
reside a representation of the object as a whole, including the features which
comprise it and the relations among them. Intermediate levels will contain
frames for the various features and the slots to be filled will contain the
tactile, visual, and amodal properties of these features. The geometric
representation will be based upon the idea of a spatial polyhedron: an object
centered guide to the exploration of the object.
The architecture will be a distributed hierarchy of knowledge-based modules,
each domain specific and informationally encapsulated. These experts will be
dedicated to the exploration for and identification of the features recognized
by the system. Each module will be responsible for filling in the slots of the
frame with which it is associated. At the lowest level of the hierarchy are
the visual and tactual perception systems. At the highest level are the two
strategists: the reasoner, responsible for hypotheses generation,
disambiguation, and culling; and the explorer, responsible for generating
sensing strategies once a goal has been formulated by the reasoner.
Advisor: Committee:
R. Bajcsy S. Lederman
R. Paul
L. Shastri
Monday, 25 November 1985
10:00 a.m.
Room To Be Announced
------------------------------
Date: 20 Nov 85 12:21:19 EST
From: PRIEDITIS@RED.RUTGERS.EDU
Subject: Seminars - ANALOGICA '85 (Rutgers)
You are cordially invited to attend
ANALOGICA '85
AT RUTGERS
Monday, December 2, 1985
8:30 AM - 6:00 PM
Hill Center for the Mathematical Sciences, Room 705
Busch Campus, Rutgers University
New Brunswick, NJ
Sponsored by
Department of Computer Science
Rutgers University, New Brunswick, NJ
and
GTE Fundamental Research Laboratories
Waltham, MA
Free and Open to the Public
Analogica '85 is a multidisciplinary seminar on analogical reasoning, bringing
together researchers from various disciplines such as artificial intelligence,
cognitive psychology, linguistics, and philosophy of science. This one-day
seminar will include eight half-hour talks and a panel discussion.
Complimentary refreshments and lunch will be served.
For more information, contact:
Armand Prieditis, prieditis@rutgers.arpa, 201-932-4273
Smadar Kedar-Cabelli, kedar-cabelli@rutgers.arpa 201-932-4648
Tom Mitchell, faculty sponsor, 201-932-4716
SCHEDULE:
8:30 - 9:00 Coffee and donuts
9:00 - 9:15 Welcome
9:15 - 9:45 Dedre Gentner, University of Illinois at Urbana-Champaign
9:45 - 10:15 Mark Burstein, BBN Laboratories
10:15 - 10:30 Coffee and donuts
10:30 - 11:00 Paul Thagard, Princeton University
11:00 - 11:30 Russell Greiner, University of Toronto
11:30 - 12:00 Smadar Kedar-Cabelli, Rutgers University
12:00 - 2:30 Luncheon
2:30 - 3:00 Lindley Darden, University of Maryland
3:00 - 3:30 Bipin Indurkhya, Boston University
3:30 - 4:00 Keith Holyoak, University of Michigan
4:00 - 4:15 Coffee and Donuts
4:15 - 5:45 Panel discussion
------------------------------
Date: Mon, 18 Nov 85 17:40:07 mst
From: yorick%nmsu.csnet@CSNET-RELAY.ARPA
Subject: Conference - Foundations of AI
WORKSHOP ON THE FOUNDATIONS OF AI
Holiday Inn, Las Cruces, New Mexico
Thursday February 6th - Saturday February 8th, 1986
Wednesday, February 5th:
6.00 pm welcome reception
9.00 - 9.30am: WELCOME AND INTRODUCTION: WILKS
9.30 - 12.30: LOGICAL/PHILOSOPHICAL FOUNDATIONS OF AI (1) (Chair: Wilks)
CHANDRASEKARAN - Paradigms in AI: An historical and contemporary perspective
HALPERN* - Turing's Test and the ideology of AI
Coffee (15 mins)
NEEDHAM - There's nothing special about AI
General Discussion (30 mins)
2.00 - 5.00pm: LOGICAL/PHILOSOPHICAL FOUNDATIONS OF AI (2) (Chair: Hayes, Pat)
DREYFUS - Traditional AI: A degenerating research program
FODOR - Why there still has to be a language of thought
Coffee (15 mins)
DENNETT - The myth of original intentionality
General Discussion (30 mins)
Friday, February 7th:
9.30 - 12.30: RELATION BETWEEN FOUNDATIONS AND PROGRAMS (1) (Chair: Sleeman)
BUNDY - What Kind of Field is AI?
McCARTHY - AI Reasoning Should be Logic with Extensions
Coffee (15 mins)
CAMPBELL* - Novelties of AI: Theories, Programs, and Rational Reconstructions
General Discussion (30 mins)
2.00 - 5.00pm: RELATION BETWEEN FOUNDATIONS AND PROGRAMS (2) (Chair: Brachman)
SPARCK-JONES - What is an experiment in AI?
NEWELL - On Comparing General Cognitive Architectures
Coffee (15 mins)
PYLYSHYN - Programs as models: Can there be a strong equivalence?
General Discussion (30 mins)
6.00 pm An evening in Mexico
Saturday, February 8th:
9.30 - 12.30: AI AND OTHER DISCIPLINES (1) (Chair: Uhr)
ARBIB - How does Brain Theory Relate to AI?
CHURCHLAND - AI and the neurosciences
Coffee (15 mins)
JOSHI - AI and Linguistics
General Discussion (30 mins)
2.00 - 5.00: AI AND OTHER DISCIPLINES (2) (Chair: Schvaneveldt)
RUMELHART - AI: What can Psychology learn?
BODEN - Has AI helped Psychology?
Coffee (15 mins)
MINSKY - AI and Cognitive Science
General Discussion (30 mins)
7.30pm: WORKSHOP BANQUET (Chair: Ortony)
Sunday, February 9th:
Depart for home, skiing etc.
Each presenter will have 30 minutes to present his/her position
followed by 15 minutes allocated for discussion.
* Indicates submitted paper.
Contact: derek@nmsu.csnet or yorick@nmsu.csnet
------------------------------
End of AIList Digest
********************