Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 108
AIList Digest Thursday, 1 May 1986 Volume 4 : Issue 108
Today's Topics:
Queries - Connection Machine Articles,
Project Description - Personality Modeling,
Techniques - String Reduction,
Expert Systems - SeRIES-PC for MS-DOS,
Representation - Shape Simulation and Recognition,
Linguistics - Xerox vs. xerox
----------------------------------------------------------------------
Date: 29 Apr 86 01:24:43 GMT
From: hplabs!sdcrdcf!usc-oberon!bacall!iketani@ucbvax.berkeley.edu
(Dana Todd Iketani)
Subject: connection machine articles
I am looking for some information about the Connection
Machine from MIT/Thinking Machines Inc. Could someone
send me some pointers for some real articles? Or an
article with a good bibliography? I've found plenty
of popular literature articles, but the only technical
paper I've found is the MIT memo by Hillis. Thanks
in advance.
d. todd Iketani
USENET: usc-cse!iketani
ARPANET: IKETANI@USC-ECL
------------------------------
Date: 29 APR 86 1404 UT
From: MCLOUH85%IRLEARN.BITNET@WISCVM.WISC.EDU
Subject: RESEARCH IN PROGRESS AT UCD IRELAND
I am doing research in AI at University College Dublin in Ireland.
In the research project we are working on we have set ourselves the
goal of developing a system which will play the part of a telephone
receptionist in "typed" telephone conversations. In our research we
have had to address ourselves to a number of problems in the areas
of planning and user modelling.
In this note I would like to say something about our work on user
modelling.
It became clear to us that in the live data which we collected
as part of the research that the receptionist seemed to have a
very rich model of the people she spoke to. These models were
much richer than any we have read about in the AI literature.
We discussed a number of the transcripts of conversations with
the receptionist and she confirmed that she was using quite detailed
models of the people she spoke to. In the vast majority of the cases
she had never met these people or spoken to them before the conversations.
It appears that quite early on in the conversation she would classify
the caller as being of a particular "type" and thereafter she would
apply any knowledge she had about that "type" of person to build a
basic model of the caller.
We have been working on applying this knowledge about stereotypes to
the system we are building and we have developed a knowledge structure
which we call a Persona. A Persona can contain knowledge about
the goals, plans, obligations, beliefs, and props which we associate with
a typical member of a class of people. It can also contain knowledge
about the types of situations in which we would normally find such people,
and the props which we might associate with them.
So far we have been working on delevoping Personae which model
common occupations, such as , the Salesman, the receptionist, the telephone
operator.
However, it is our intention in the future to try and develop Personae
which model particular attributes such as Friendly, Aggressive etc and to
see if we can develop ways of combining these to produce personality
models.
Well, having given an overview of the work we are doing, I would be
grateful if there are any readers who share our interest in Personality
modelling and stereotypes who might have any comments to make or
who might be able to reccommend particular papers in the area which
might be of interest to us.
Henry B McLoughlin.
Department of Computer Science
University College Dublin
MCLOUH85@IRLEARN
------------------------------
Date: 24 Apr 86 13:26:37 GMT
From: uwvax!harvard!cmcl2!philabs!linus!security!jkm@ucbvax.berkeley.edu
(Jonathan K.Millen)
Subject: Re: String reduction
If you are interested in an applicative Lisp-like language
based on string substitution and reduction, you might want to look
at "TRAC, A Procedure-Describing Language for the Reactive Typewriter",
by Calvin N. Mooers, Comm ACM, Vol. 9, No. 3, March, 1966.
Jon Millen
------------------------------
Date: Tue 29 Apr 86 09:41:32-PDT
From: Lou Fried <FRIED@SRI-KL.ARPA>
Subject: Expert Systems for MS-DOS
Please include:
SeRIES-PC, language IQLISP, $5,000
SRI International
333 Ravenswood Avenue
Menlo Park, CA 94025
Contact: Bob Wohlsen, x 4408
------------------------------
Date: Tue, 29 Apr 86 15:47 EDT
From: Seth Steinberg <sas@BBN-VAX.ARPA>
Subject: Shape -- Simulation and Recognition
This hooks, rings and shapes discussion points at that AI contains a
lot of simulation. One useful model of AI programming is to view a
program as an intertwining of simulation and recognition, especially if
you are willing to think of these concepts a bit more generally than
ordinarily.
- A game playing program will play the game forwards (simulation) and
then choose a course of play to follow (recognition).
- A logic programming system will follow the ramifications of an
assertion by forward chaining (simulation) and then seek a particular
fact in the rule base (recognition).
- A robotics program will examine its goals and its model of the world
(recognition) and then test if a pariticular motion is useful or
possible (simulation). (Who was it who wrote in saying that all he
needed was an AT function and everything else would be easy?)
[That was a quote from Peter Cheeseman's early days. -- KIL]
This is not a strict breakdown, but rather a useful insight which
explains why certain problems are solved in certain ways. Thinking of
part of the program as the simulator and part as the recognizer can
reveal two of the conflicting forces in the resolution of the problem.
No matter how they are described, there are two fundamental dynamic
elements which force the tradeoffs required to engineer a working
program.
- Anoop Gupta who is working on implementing OPS5-like rule resolution
systems on parallel machines noted that programs broke into Problem
Space Search which grows exponentially (simulation) which is where the
parallelism comes from and Knowledge Space Search which applies the
programs knowledge of the domain to restrict the growth rate of
parallelism (recognition).
- A number of other researchers have broken problem solving into the
marshalling of alternatives (simulation) followed by the focussing of
attention on the most promising (recognition). Phil Agre argues that
one purpose of consciousness is to control the direction of attention
to the alternatives.
My personal feeling is that AI isn't going to tackled everyday
knowledge until it starts simulating everyday things. Steamer, the
large steam engine expert contained a gigantic Fortran program
(rewritten in Lisp) to simulate the engine. I don't know of any
programs that simulate a kitchen. AI has already borrowed a lot of
object oriented programming from Simula which is a simulation language.
Maybe AI programmers, being forced to deal with the problems of
simulation will find other as yet neglected tools.
Seth Steinberg
------------------------------
Date: Tue, 29 Apr 86 11:12:48 cdt
From: bulko@SALLY.UTEXAS.EDU (Bill Bulko)
Reply-to: bulko@sally.UUCP (Bill Bulko)
Subject: Re: do not "xerox" this message
I don't see what the big deal is about using "xerox" as a verb. I often
hear "Coke" used to mean "[virtually any] carbonated soft drink". How
often do you order a Coke at a fast-food place and get Pepsi or RC?
If anything, the use of "xerox" as a verb is a tribute to the
contribution Xerox has made to photocopying.
Bill
_______________________________________________________________________________
"In the knowledge lies the power." -- Edward A. Feigenbaum
"Knowledge is good." -- Emil Faber
Bill Bulko The University of Texas
bulko@sally.UTEXAS.EDU Department of Computer Sciences
------------------------------
Date: Tue 29 Apr 86 11:40:41-PDT
From: Pat Hayes <PHayes@SRI-KL>
Subject: Trademarks
Laws concerning trademark usage aside, De Smedt is perfectly correct
in pointing out that the verb 'to xerox', meaning to copy on a
dry-xerographic copier, and associated constructions ( a xerox copy,
etc. ), are now in fact part of the language. The distinction between
'xerox' and 'Xerox' seems quite clear, and it might be more sensible
for the company to insist on 'correct' usage of the latter rather than
the former. Its no use, guys, you can't stop people using the word in
the way they want to. A dictionary which omitted 'to xerox' would not
be accurate. Its an inevitable consequence of the fact that for many
years, all the copiers WERE Xerox machines, just like Bics and Kodaks.
One can't have no competitors while a new technology is entering the
marketplace, and expect not to be identified with it. Especially if
one has also invented a neat, original, snappy name for it ( like Bic
and Kodak ). Its the price of success.
Pat Hayes
Schlumberger ( a word which will never enter the language )
------------------------------
End of AIList Digest
********************
