Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 229
AIList Digest Thursday, 23 Oct 1986 Volume 4 : Issue 229
Today's Topics:
Seminars - Toward a Learning Robot (CMU) &
Implementing Scheme on a Personal Computer (SMU) &
Functional Representations in Knowledge Programming (UTexas) &
The Hypotheses Underlying Connectionism (UCB) &
Advances in Computational Robotics (CMU) &
More Agents are Better Than One (SU) &
Automatic Schematics Drafting (CMU) &
Learning by Failing to Explain (MIT)
----------------------------------------------------------------------
Date: 15 October 1986 1505-EDT
From: Elaine Atkinson@A.CS.CMU.EDU
Subject: Seminar - Toward a Learning Robot (CMU)
SPEAKER: Tom Mitchell, CMU, CS Department
TITLE: "Toward a Learning Robot"
DATE: Thursday, October 16
TIME: 4:00 p.m.
PLACE: Adamson Wing, Baker Hall
ABSTRACT: Consider the problem of constructing a learning robot; that
is, a system that interfaces to some environment via a set of sensors
and effectors, and which builds up a theory of its environment in order
to control the environment in accordance with its goals. One
instantiation of this problem is to construct a hand-eye system that
can learn to manipulate a collection of blocks and to build simple
structures from these blocks.
We are starting a new research project to develop such a learning
robot, and this talk will present some preliminary ideas about how
to proceed. The talk will consider a number of questions such as
what general cognitive architecture seems reasonable? What kinds
of knowledge must such a robot learn? How should this knowledge
be represented? How will it learn? How can the robot solve
problems with only an incomplete understanding of its world? Can
it use sensory feedback to make up for ambiguity in its world
theory? There will probably be more questions than answers,
so please bring your own.
------------------------------
Date: WED, 20 apr 86 17:02:23 CDT
From: leff%smu@csnet-relay
Subject: Seminar - Implementing Scheme on a Personal Computer (SMU)
Implementing Scheme on a Personal Computer
Speaker: David Bartley Location: 315 SIC
Texas Instruments Time: 2:00 PM
PC Scheme is an implementation of the Scheme language, a lexically scoped,
applicative order, and properly tail-recursive dialect of LISP. PC Scheme was
implemented for IBM and TI personal computers within the Symbolic Computing
Laboratory at Texas Instruments. The presentation will examine some of the
pragmatic aspects of developing a production-quality LISP system for small
machines. These include: compilation vs. interpretation, using a byte-threaded
virtual machine for compact code, the architecture of the virtual machine,
runtime representation issues, compiler design, debugging issues, and
performance. Some significant differences between LISP and conventional
language implementations will be highlighted.
------------------------------
Date: Fri 17 Oct 86 12:56:46-CDT
From: Ellie Huck <AI.ELLIE@MCC.COM>
Subject: Seminar - Functional Representations in Knowledge Programming (UTexas)
Please join the AI Group for the following talk October 22 at 11:00am
in the Balcones 4th Floor Conference Room 4.302:
KNOWLEDGE PROGRAMMING USING FUNCTIONAL REPRESENTATIONS
Peter E. Hart
Syntelligence
SYNTEL is a novel knowledge representation language that provides
traditional features of expert system shells within a pure functional
programming paradigm. However, it differs sharply from existing
functional languages in many ways, ranging from its ability to deal
with uncertainty to its evaluation procedures. A very flexible
user-interface facility, tightly integrated with the SYNTEL
interpreter, gives the knowledge engineer full control over both form
and content of the end-user system. SYNTEL executes in both LISP
machine and IBM mainframe/workstation environments, and has been used
to develop large knowledge bases dealing with the assessment of
financial risks. This talk will present an overview of its
architecture, as well as describe the real-world problems that
motivated its development.
October 22, 1986
11:00am
Balcones Room 4.302
------------------------------
Date: Mon, 20 Oct 86 10:30:58 PDT
From: admin%cogsci.Berkeley.EDU@berkeley.edu (Cognitive Science Program)
Subject: Seminar - The Hypotheses Underlying Connectionism (UCB)
BERKELEY COGNITIVE SCIENCE PROGRAM
Cognitive Science Seminar - IDS 237A
Tuesday, October 28, 11:00 - 12:30*
2515 Tolman Hall
Discussion: 12:30 - 1:30
2515 Tolman Hall
``The Hypotheses Underlying Connectionism''
Paul Smolensky
Department of Computer Science & Institute of Cognitive Science
University of Colorado at Boulder
Cognitive models using massively parallel, nonsymbolic computation
have now been developed for a considerable variety of cognitive
processes. What are the essential hypotheses underlying these
connectionist models? A satisfactory formulation of these hy-
potheses must handle a number of attacks:
-Nothing really new can be offered since Turing machines are universal
-Connectionism just offers implementation details
-Conscious, rule-guided behavior is ignored
-The wrong kind of explanations are given for behavior
-The models are too neurally unfaithful
-Logic, rationality, and the structure of mental states are ignored
-Useful AI concepts like frames and productions are ignored.
Firstly, an introduction to connectionist models which describes
the kind of computation they use will be presented and secondly,
a general connectionist approach that faces the challenges listed
above will be introduced.
------------------------------
Date: 20 October 1986 1309-EDT
From: Richard Wallstein@A.CS.CMU.EDU
Subject: Seminar - Advances in Computational Robotics (CMU)
Robotics Seminar, FRIDAY Oct. 24, 2 PM, 4623 WeH
John H. Reif
Computer Science Department
Duke University
ADVANCES IN THE THEORY OF COMPUTATIONAL ROBOTICS
This talk surveys work on the computational complexity of various movement
planning problems relevant to robotics. The generalized mover's problem is to
plan a sequence of movements of linked polyhedra through 3-dimensional
Euclidean space, avoiding contact with a fixed set of polyhedra obstacles.
We discuss algorithms for solving restricted mover's problems and our proof
that generalized mover's problems are polynominal space hard.
We also discuss our results on the computational complexity (both algorithms
and lower bounds) of three other quite different types of movement problems;
1. movement planning in the presence of friction;
2. minimal movement planning;
3. dynamic movement planning with moving obstacles.
------------------------------
Date: 20 Oct 86 1141 PDT
From: Vladimir Lifschitz <VAL@SAIL.STANFORD.EDU>
Subject: Seminar - More Agents are Better Than One (SU)
MORE AGENTS ARE BETTER THAN ONE
Michael Georgeff
Artificial Intelligence Center
SRI International
Thursday, October 23, 4pm
MJH 252
A recent paper by Steve Hanks and Drew Mcdermott shows how some
previous "solutions" to the frame problem turn out to be inadequate,
despite appearances otherwise. They use a simple example -- come to
be called the "Yale Shooting Problem" -- for which it is impossible to
derive some expected results -- in this case, that the target of a
shooting event ceases living. Such difficulties, they suggest, call
into question the utility of nonmonotonic logics for solving the frame
problem.
In this talk, we describe a theory of action suited to multiagent
domains, and show how this formulation avoids the problems raised by
Hanks and McDermott. In particular, we show how the Yale Shooting
Problem can be solved using a generalized form of the situation
calculus for multiagent domains, together with notions of causality
and independence. The solution does not rely on complex
generalizations of nonmonotonic logics or circumscription, but instead
uses traditional circumscription. We will also argue that most
problems traditionally viewed as involving a single agent are better
formulated as multiagent problems, and that the frame problem, as
usually posed, is not what we should be attempting to solve.
------------------------------
Date: 20 Oct 86 15:23:35 EDT
From: Steven.Minton@k.cs.cmu.edu
Subject: Seminar - Automatic Schematics Drafting (CMU)
This week's seminar is being led by Raul Valdes-Perez. Friday, 3:15
in 7220. Be there. Here's the abstract:
Title: "Automatic Schematics Drafting: Aesthetic Configuration
as a Design Task".
To draft a schematic means to depict (say on paper) the
electrical connections and function of a circuit.
Aspects of this work are the following:
1. A design task that uses other than a production-system architecture.
2. An approach to "space planning" that is modern in the sense of exploiting
dependency-directed backtracking and constraint-posting.
3. The idea of contradicton-fixing rules that exploit the richness of
information when an inconsistency occurs.
4. Study of a linear-inequality-based representation of partial task
solutions, and the properties of this representation.
5. A backtracking scheme suited to the search regimen used.
------------------------------
Date: Tue, 21 Oct 1986 12:05 EDT
From: JHC%OZ.AI.MIT.EDU@XX.LCS.MIT.EDU
Subject: Seminar - Learning by Failing to Explain (MIT)
LEARNING BY FAILING TO EXPLAIN
Robert Joseph Hall
MIT Artificial Intelligence Laboratory
Explanation-based Generalization depends on having an explanation on
which to base generalization. Thus, a system with an incomplete or
intractable explanatory mechanism will not be able to generalize some
examples. It is not necessary, in those cases, to give up and resort
to purely empirical generalization methods, because the system may
already know almost everything it needs to explain the precedent.
Learning by Failing to Explain is a method which exploits current
knowledge to prune complex precedents and rules, isolating their
mysterious parts. This paper describes two techniques for Learning by
Failing to Explain: Precedent Analysis, partial analysis of a
precedent or rule to isolate the mysterious new technique(s) it
embodies; and Rule Re-analysis, re-analyzing old rules in terms of new
rules to obtain a more general set.
Thursday, October 23, 4pm
NE-43, 8th floor playroom
------------------------------
End of AIList Digest
********************
