Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 116
AIList Digest Thursday, 8 May 1986 Volume 4 : Issue 116
Today's Topics:
Seminars - Planning, Knowledge, and Action (UPenn) &
Knowledge Engineering as Ontological Analysis (SU) &
Default Theories and Autoepistemic Logic (CSLI) &
NL Database Query Systems (UPenn) &
Sequential and Parallel Inference Machines (Edinburgh) &
Ulysses Expert-System VLSI Design Environment (UPenn) &
Granularity (SRI) &
Eazyflow: an Effective Alternative to Dataflow (CMU),
Conference - Workshop on Intelligent Interfaces at AAAI-86
----------------------------------------------------------------------
Date: Mon, 5 May 86 13:56 EDT
From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - Planning, Knowledge, and Action (UPenn)
Colloquium - University of Pennsylvania
3:00pm May 6, 1986
216 Moore School
A FIRST ORDER THEORY OF PLANNING, KNOWLEDGE, AND ACTION
LEORA MORGENSTERN - NEW YORK UNIVERSITY
Most AI planners work on the assumption that they have complete knowledge of
their problem domain and situation, so that formulating a plan consists of
searching through some pre-packaged list of action operators for an action
sequence that achieves some desired goal. Real life planning rarely works this
way, because we usually don't have enough information to map out a detailed
plan of action when we start out. Instead, we initially draw up a sketchy plan
and fill in details as we proceed and gain more exact information about the
world.
This talk will present a formalism that is expressive enough to describe this
flexible planning process. We begin by discussing the various requirements
that such a formalism must meet, and present a syntactic theory of knowledge
that meets these requirements. Next, we discuss the paradoxes, such as the
Knower Paradox, that arise from syntactic treatments of knowledge, and propose
a solution to these paradoxes based on Kripke's solution to the Liar Paradox.
Finally, we give solutions to the Knowledge Preconditions and Ignorant Agent
Problems as part of an integrated theory of planning.
The talk will include comparisons of our theory with other syntactic and modal
theories such as Konolige's and Moore's. We will demonstrate that our theory
is powerful enough to solve classes of problems that these theories cannot
handle.
------------------------------
Date: Mon 5 May 86 17:20:08-PDT
From: Christine Pasley <pasley@SRI-KL>
Subject: Seminar - Knowledge Engineering as Ontological Analysis (SU)
CS529 - AI In Design & Manufacturing
Instructor: Dr. J. M. Tenenbaum
Title: Knowledge Engineering as Ontological Analysis
Speaker: Patrick Hayes
From: Schlumberger Palo Alto Research
Date: Wednesday, May 7, 1986
Time: 4:00 - 5:30
Place: Terman 556
When designing a knowledge-base for use by an AI system, it is important to
bear in mind how utterly stupid computers are. We must provide them with a
vocabulary in which to think about their world, and the scope of their thoughts
is then limited by the expressiveness of this vocabulay: in particular, the
kinds of object it is able to talk about. This talk will illustrate this
point, and emphasise how important it is to choose the representational
vocabulary to fit both the limitations and the range of the systems desired
abilities. Ways of referring to times and events will be used as examples.
Visitors welcome!
------------------------------
Date: 05 May 86 1625 PDT
From: Vladimir Lifschitz <VAL@SU-AI.ARPA>
Subject: Seminar - Default Theories and Autoepistemic Logic (CSLI)
ON THE RELATION BETWEEN DEFAULT THEORIES AND AUTOEPISTEMIC LOGIC
Kurt Konolige
SRI International and CSLI
Common Sense and Non-Monotonic Reasoning Seminar
Thursday, May 8, 4pm
MJH 252
Default theories are a formal means of reasoning about defaults: what
normally is the case, in the absence of contradicting information.
Autoepistemic theories, on the other hand, are meant to describe the
consequences of reasoning about ignorance: what must be true if a
certain fact is not known. Although the motivation and formal
character of these systems are different, a closer analysis shows that
they bear a common trait, which is the indexical nature of certain
elements in the theory. In this talk I will show how default theories
can be reanalyzed as a restricted type of indexical theory. The
benefits of this analysis are that it gives a clear (and clearly
intuitive) semantics to default theories, and combines the expressive
power of default and autoepistemic logics in a single framework.
------------------------------
Date: Tue, 6 May 86 21:40 EDT
From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - NL Database Query Systems (UPenn)
Forwarded From: Naoki Abe <Abe@UPenn> on Tue 6 May 1986 at 18:18
A REMINDER OF A COLLOQUIUM
Thursday 5/8, 3:00pm, 216 Moore School
There will be an interesting talk on natural language database query systems
by Dr. Stanley R. Petrick of I.B.M. Thomas J. Watson Research Center. Dr.
Petrick is a former president of the Association for Computational
Linguistics, and is known for developing the first parsing algorithm for
transformational grammars, characterizing various parsing algorithms for
context free grammars in terms of push down automata, as well as his earlier
work on the minimal covering problem and its application on speech
recognition. In this talk he will discuss more practical issues concerning
natural language query systems. The following is the abstract of this talk.
Natural Language Database Query Systems
Dr. Stanley R. Petrick
Thomas J. Watson Research Center, I.B.M.
In recent years many computer systems have been developed with limited
capabilities for understanding natural language requests for information
from a given database and for responding appropriately. In this talk we
shall attempt to characterize the theory underlying these systems and the
level of performance that they have demonstrated. Special attention will be
given to the problem of customizing such systems to handle new databases.
Illustrative material will be drawn from the T.Q.A. (Transformational
Question Answering) system, an experimental prototype being developed at
I.B.M. Research.
------------------------------
Date: Tue, 6 May 86 10:18:52 -0100
From: Gideon Sahar <gideon%edai.edinburgh.ac.uk@Cs.Ucl.AC.UK>
Subject: Seminar - Sequential and Parallel Inference Machines (Edinburgh)
EDINBURGH AI SEMINARS
Date: Wednesday, 7th May l986
Time: 2.00 p.m.
Place: Department of Artificial Intelligence
Seminar Room - F10
80 South Bridge
EDINBURGH.
Professor David H.D. Warren, Department of Computer Science, University
of Manchester will give a seminar entitled - "Sequential and Parallel
Inference Machines".
There is a growing interest, stimulated in large part by Japan's Fifth
Generation project, in computer architectures where the basic machine
language is a form of symbolic logic, and the basic machine operation
is a form of logical inference. Prolog is the best known, but not the
only, example of such a language.
How fast can such machines run? I will consider both sequential
machines, which perform only one logical inference at a time, and
parallel machines, which can perform more than one logical inference at
a time.
First, I will describe my work with Evan Tick on the design of a Prolog
instruction set and pipelined processor. This work suggests that a
sequential Prolog machine can achieve a speed approaching one million
logical inferences per second (IM LIPS) with current device technology.
This estimate has been confirmed by experimental prototypes, produced
at Berkeley and NEC.
In the second part of the talk I will discuss various approaches to
exploiting parallelism, including the Argonne approach to
or-parallelism, and the approach of DeGroot and others to
and-parallelism.
------------------------------
Date: Wed, 7 May 86 14:31 EDT
From: Tim Finin <Tim%upenn.csnet@CSNET-RELAY.ARPA>
Subject: Seminar - Ulysses Expert-System VLSI Design Environment (UPenn)
Electrical Engineering Colloquium - University of Pennsylvania
11:00am May 9, 1986 - 129 Pender Lab
Ulysses -- An Expert-System Based VLSI Design Environment
Michael l. Bushnell
Carnegie-Mellon University
It has recently been observed that the initial engineering design cost
for VLSI circuits is beginning to exceed the lifetime production cost.
In order to reduce this prohibitive design cost, which is limiting the
practical applications of VLSI technology, we need a real increase in
the automation of design activities. Ulysses is a VLSI CAD environment
which effectively addresses the problem of CAD tool integration and
which also allows further automation of the VLSI design process. The
goal of this environment is to raise the designer interface for CAD
systems from the CAD tool level to the design task level. The environment
is intended to be used in design synthesis, design-for-testability,
analysis, verification and optimization activities at all levels of
VLSI design. Specifically, Ulysses alleviates the problems caused by
incompatible file formats for CAD tools, allows one to codify a design
methodology, allows the methodology to be semi-automatically executed
and allows the VLSI design space to be explicitly represented. The
environment automatically executes existing CAD tools, according to
instructions expressed in the codified design methodology, in order to
accomplish design tasks. Ulysses keeps track of the progress of a
design and lets the designer explore the design space. Ulysses uses
Artificial Intelligence methods, functions as an interactive expert
system, and interprtets language descriptions of design tasks, which
are described in the Scripts language. Alternatively, the Scripts
language may be viewed as an organization-structuring language for CAD
applications in engineering. An example of an IC layout design task
will be presented, in which a knowledge-based router, a layout
synthesizer, and an interactive floor planner will be controlled by
Ulysses in a non-deterministic and opportunistic fashion in order to
produce a viable IC layout from a circuit description expressed in the
logic element/transistor level in a hardware description language.
------------------------------
Date: Wed 7 May 86 14:51:37-PDT
From: LANSKY@SRI-AI.ARPA
Subject: Seminar - Granularity (SRI)
GRANULARITY
Jerry R. Hobbs (HOBBS@SRI-AI)
Artificial Intelligence Center, SRI International
CSLI, Stanford University
11:00 AM, MONDAY, May 12
SRI International, Building E, Room EJ228 (new conference room)
We look at the world under various grain sizes and abstract from it only
those things that serve our present interests. We can view a road,for
example, as a line, a surface, or a volume. Such abstractions enable us
to reason about situations without getting lost in irrelevant
complexities. Knowledge-rich intelligent systems will have to have
similar capabilities. In this talk I will present a framework in which
we can understand such systems. In this framework, a knowledge base
consists of a global theory together with a large number of relatively
simple, idealized, grain-dependent local theories, interrelated by
articulation axioms. In a complex situation, the crucial features are
abstracted from the environment, determining a granularity, and the
corresponding local theory is selected. This is the only computation
done in the global theory. The local theory is then applied in the bulk
of the problem-solving process. When shifts in perspective are
required, articulation axioms are used to translate the problem and
partial results from one local theory to another. In terms of this
framework, I will discuss idealization, the concepts of supervenience
and reducibility, prototype-deformation types of description, and the
emergence of global properties from local phenomena, and the
relationship of granularity to circumscription. Several examples of
uses of this framework from a wide variety of applications will be
given.
VISITORS: Please arrive 5 minutes early so that you can be escorted up
from the E-building receptionist's desk. Thanks!
------------------------------
Date: 6 May 1986 0754-EDT
From: Theona Stefanis@A.CS.CMU.EDU
Subject: Seminar - Eazyflow: an Effective Alternative to Dataflow (CMU)
PS SEMINAR
Edward Ashcroft, SRI
Date: Monday, 12 May
Time: 3:30
Place: WeH 5409
Title: "Eazyflow: an Effective Alternative to Dataflow"
Eazyflow is a evaluation strategy for Operator Nets. Syntactically,
operator nets are similar to dataflow graphs. Their semantics is
expressed mathematically, and is more general and elegant than the
semantics of dataflow networks. Various ways of specifying their
operational semantics are possible, and eazyflow is one such way, that
is a hybrid of demand-driven and data-driven evaluation. (Data-driven
evaluation is what is normally called dataflow. Demand-driven
evaluation avoids a lot of the problems that dataflow has. Eazyflow is
basically demand-driven, with data-driven computation taking place when
it can do so without causing too many problems.)
This talk will
describe operator nets and their mathematical semantics, indicate how
they correspond exactly to programs in the language Lucid, show how
eazyflow is often superior to dataflow, and briefly describe the
architecture of the eazyflow engine that is soon to be built at SRI (the
Eazyflow Architecture Project is currently part of the DARPA Strategic
Computing Program). Also, some simulation results that have been
obtained for the architecture will be described.
------------------------------
Date: 6 May 1986 12:02-PDT
From: Neches@isi-vaxa.arpa (Robert Neches)
Subject: Conference - Workshop on Intelligent Interfaces at AAAI-86
A workshop on Intelligent Interfaces is scheduled to be held on Thursday,
August 13th, as part of the AAAI conference in Philadelphia. We would
like to bring the call for abstracts to your attention, and would appreciate
it if you would circulate it to anyone else who might find it of interest.
-- Tom Kaczmarek (Kaczmarek@USC-ISIB.arpa)
Bob Neches (Neches@ISI-Vaxa.arpa)
Workshop Co-chairs.
USC / Information Sciences Institute
4676 Admiralty Way
Marina del Rey, CA 90292
(213) 822-1511
Questions may be addressed to either chairman; abstracts should be sent to
Tom Kaczmarek by June 15. The complete call for abstracts follows.
*****************************************************************************
WORKSHOP ON INTELLIGENT INTERFACES AT AAAI-86
Many AI techniques are applicable to building better human-machine
interfaces. The purpose of this workshop is to investigate intelligent
interface techniques that can potentially span many interaction modalities.
The workshop will discuss interfaces to knowledge-based systems as well
conventional interactive systems. Past work in this area has been directed
at using AI to provide either an "intelligent apprentice" or a collection of
"power tools." The intelligent apprentice emphasizes assistance based on an
understanding of the user's intentions and task domain. The power tool
approach emphasizes a powerful command set, but leaves the responsibility
for selecting and applying commands in the hands of the user. This workshop
is concerned not just with the extremes of this dichotomy, but also with
work that shows how to blend the two approaches effectively. Work on
specific media and modalities, (e.g., natural language text or speech
understanding) is also relevant in that it can provide abstractions of
understanding and generation that will be potentailly useful across a wide
range of interface media and modalities.
Topics to be discussed:
What are the fundamental interface problems that AI can help solve?
What specific AI techniques can be useful in solving these problems?
What abstractions of "understanding" and "generation" can come from
work on natural language text and speech?
What are the possibilities for symbiotic relationships between
intelligent interfaces and intelligent systems?
What does it take to create intelligent interfaces to conventional
interactive systems?
Are the power tools and intelligent assistance approaches at odds
with one another? Are middle-of-the-road approaches
motivated by pragmatism or principle?
Organizers: The workshop organizers are Thomas Kaczmarek, Larry Miller,
Robert Neches and Norman Sondheimer of the USC/Information Sciences
Institute.
Participation: The workshop will run for a full day on Thursday, August 13
at the University of Pennsylvania. The format will be a combination of
short informal presentations and open discussions with the former being used
to stimulate the latter. These will be organized in four sessions, the
topics of which will be finalized after reviewing the declared interests of
participants. Attendence will be by invitation only; there will be a
maximum of 50 participants. Those wishing to participate should submit four
copies of a 1000-word abstract describing either their work building
intelligent interfaces or a position on a topic relevant to the goals of the
workshop. Abstracts should provide contact information at the top, as they
will be duplicated and distributed to the other workshop attendees.
Participants with a willingness to make a short presentation (15-30 minutes)
about either their research or a position on a relevant topic should
indicate this desire in a cover letter sent with the abstract. If multiple
members of a research group would like to attend, please indicate the
number involved in the cover letter also. Abstracts should be sent to
Thomas Kaczmarek, USC/ISI, 4676 Admiralty Way, Marina del Rey, CA.
90292-6695. They may also be transmitted electronically to
Kaczmarek@USC-ISIB.arpa. The deadline for submission of abstracts is June
15, 1986. Invitations will be issued by July 15.
------------------------------
End of AIList Digest
********************
