LAWS@SRI-AI.ARPA (02/26/85)
From: AIList Moderator Kenneth Laws <AIList-REQUEST@SRI-AI.ARPA>
AIList Digest Tuesday, 26 Feb 1985 Volume 3 : Issue 26
Today's Topics:
Linguistics - Wally,
Bindings - General Research Corp.,
Publications - Request for Sources & TARGET AI Newsletter,
AI Tools - YLISP & KayPro AI Languages,
News - Recent Articles,
Humor - EURISKO & Programming the User-Friendly Dog,
Seminar - Motion Planning with Uncertainty (MIT)
----------------------------------------------------------------------
Date: Sunday, 24-Feb-85 18:24:20-GMT
From: GORDON JOLY (on ERCC DEC-10) <GCJ%edxa@ucl-cs.arpa>
Subject: Wally.
There was some discussion recently in the Guardian newspaper here
in the U.K. about the word `wally'. Does the word exist on the other
side of the Atlantic (or elsewhere) and if so what meaning does it have?
Gordon Joly
gcj%edxa@ucl-cs.arpa
School of Mathematical Sciences,
Queen Mary College, Mile End Road,
LONDON E1 4NS, UK.
------------------------------
Date: Fri 22 Feb 85 15:41:04-PST
From: Rene Bach <BACH@SU-SCORE.ARPA>
Subject: General Research Corp. pointer request
Does someone know the address/phone number of the company ? They apparently
have developed an expert system to tune a VAX VMS operating system we might
be interested in using. It is build on TIMM and is called TUNER.
Thanks for any info.
Rene Bach, Varian Associates
Bach@score
------------------------------
Date: 22 Feb 85 08:42:26 GMT (Friday)
From: Martin Cooper <Cooper.rx@XEROX.ARPA>
Subject: Re: Slightly Depressed....
I certainly feel the same sense of deprivation, since as far as I'm
concerned, all the seminars are on the wrong side of the Atlantic, and a
very long way from the UK.
I wonder if it would be possible for the existence of related papers
and/or recordings to be mentioned along with the seminar announcements,
or in a related message on this list.
Martin.
------------------------------
Date: Fri 22 Feb 85 10:36:44-PST
From: Ted Markowitz <G.TJM@SU-SCORE.ARPA>
Subject: AI newsletter and request for recommendations
I just received a flier from another AI newsletter called 'TARGET
The AI Business Newletter'. In 1985 it promises
Analysis of installed base of AI and standard machines
used for symbolic processing.
The market for natural language.
AI in the micro marketplace.
AAAI/IJCAI coverage.
Outlook on venture capital and corporate funding.
Price: $190/year
Where: Target Technologies, 3000 Sand Hill Rd., Bldg 1, Suite 255
Menlo Park, CA 94025.
--ted
PS: Is there a consensus in the group for which of these 'digests'
is really worth the several hundred bucks/yr.? I'd like to get
my folks to order one, but I want to make sure it's worth
something.
------------------------------
Date: Sun, 24 Feb 85 13:23:02 -0200
From: jaakov%wisdom.BITNET@WISCVM.ARPA (Jacob Levy)
Subject: YLISP available for FTP
Hi!
I am pleased to announce that YLISP is finally available for FTP
from 'maryland.arpa' using user id 'ftp' and password 'anonymous'. The
stuff is in directory YLISP (note capitals); There are 20 files in all
to copy - 16 files containing the system, named FTP_Y.[1-16], a makefile
a recursive copying program named 'copy' and 2 READ.ME files. Make sure
that the receiving system has ~4500 K disk space on the file system you
copy it to.
Before trying to install the system, please read 'READ.ME.FTP'
carefully. All bugs, complaints, requests and suggestions please mail to
BITNET: jaakov@wisdom
CSNET and ARPA: jaakov%wisdom.bitnet@wiscvm.ARPA
UUCP: (if all else fails..) ..!decvax!humus!wisdom!jaakov
POSTAL: Jacob Levy
Dept of Applied Math,
Weizmann Institute of Science,
Rehovot 71600, ISRAEL
PS - The system will be available on BITNET pretty soon also. Separate
anouncement to BITNET users will follow.
------------------------------
Date: 22 Feb 1985 17:09:06 EST
From: BASKEYFIELDM@USC-ISI.ARPA
Subject: AI Info.
I would like to know if anyone out there has an AI language that can
be run on a KAYPRO 4/84 using the CP/M operating system?
Recent articles:
There is an interesting and introductory article article in the
March 1985 issue of COMPUTERS AND ELECTRONICS, pp. 69-73, entitled
"Expert Systems on Microcomputers." The article describes the very
basics of decision-support systems and how AI fits in. The systems
covered in the article are M.1 and M.1a (Teknolledge), Expert Ease
(Expert Systems), and MVP-Forth (Mountain View Press). These three
systems all run on the IBM PC/XT.
In addition, there is an article "AI On A Chip" and another, "FORTH
and AI" which prove interestign reading for new people in the AI
field.
Mark Baskeyfield
Naval Postgraduate School
Monterey, CA 93943
baskeyfieldm@usc-isia.arpa
Thanks!!
------------------------------
Date: Thu, 21 Feb 85 09:27:34 cst
From: Laurence Leff <leff%smu.csnet@csnet-relay.arpa>
Subject: Recent Articles
Infoworld February 25, 1985 Page 5 Editorial
'The next big lemminglike rush will be to artificial intelligence. AI will
be the most despised and abused [software concept] of the next year. So
in a perverse way, AI is an exciting opportunity for people who recognize
what it can do for customers.'
Mitch Kapor, Chairman of Lotus Development
Some people are avoiding the AI label due to AI-hype and Rube Goldberg
overdesign. Microsoft's Bill Gates has used the term "softer software"
instead of AI. This is for systems that will learn the user's work
patterns and help execute them.
For example, if a person dials up his mainframe, gets some data, does
some spread sheet processing and generates some graphics pasting it into
a report, the system should figure out that is his pattern and start
doing it automatically. Furthermore it should handle a request on
Friday like "generate the usual sales report but give me a separate
graph on what is happening in Europe"
Also software should determine which configuration a user has so he does
not have to enter information as to what graphics card and printer he
is using.
------------------------------
Date: Mon 25 Feb 85 10:48:43-PST
From: Ken Laws <Laws@SRI-AI.ARPA>
Subject: Recent Articles
IEEE Computer Graphics and Applications, February 1985:
From Paper Drawings to Computer-Aided Design, by M. Karima, K. Sadhal,
and T. McNeil of Atomic Energy of Canada Limited, pp. 27-39:
A survey of techniques and difficulties in automated optical
entry of complex schematics and wiring diagrams. Thirty papers
are referenced.
An Overview of Analytic Solid Modeling, by M. Casale and E. Stanton
of PDA Engineering, pp. 45-56:
Describes a way of modeling complex shapes as composites of parametric
curves and solids (e.g., swept or deformed regular solids). This seems
to combine the volumetric simplicity of CSG representations with the
flexibility of B-rep. Derivation of mass properties is very simple,
although the "inverse problem" of determining whether a point is inside
or outside a volume is a little complex. The ASM approach maintains
a parametric coordinate system so that physical properties (e.g.,
temperature, stress, curvature, color) can be attached to each point
on or in a solid. This makes the technique ideal for finite element
analysis.
Braintrain Seeks Educational Software from Independent Authors, p. 82:
An example is shown of this company's iconic programming language for
the ChipWits computer-graphics robots. Apparently it's a high-level
flowchart language. The simulated robots can be pitted against
various simulated environments. The game software is available for
the Macintosh and Apple II.
Badler Becomes Associate Editor-in-Chief of IEEE CG&A, p. 86:
Norman Badler is interested in artificial intelligence, particularly
as applied to simulation of human motion, so the magazine will no
doubt continue its coverage of AI-related topics.
High Technology, March 1985:
Software Tools Speed Expert System Development, by P. Kinnucan, pp. 16-20:
Describes commercially-available expert system shells, particularly KEE.
Mentions KEE for frame-based representation and for forward chaining
(as well as the usual backward chaining), ART for hypothetical and
multiworld reasoning, and Insight for database access and low price
($95, but without the ability to use variables in rules). Also mentions
M.1, LOOPS, TI's Personal Consultant, Arbie, IN-ATE, and REVEAL.
Expert System Shells Boost A.I. Market, by M. Foley, p. 21:
Further discussion of the same material.
IntelliCorp: The Selling of Artificial Intelligence, by E. Linden, pp. 22-25:
A two-page history and description of the company making KEE.
Prospecting from the Skies, by G. Graff, pp. 49-56:
Interesting discussion of the advances that can be expected soon in
remote sensing >>without<< the use of AI (but with high-resolution
multispectral data and sophisticated location- and time-specific
analysis).
Personal Robots Face Software Challenge, by M. Higgins, pp. 71-73:
Describes the primitive state of personal-robot software.
-- Ken Laws
------------------------------
Date: Thu, 21 Feb 85 19:51:37 est
From: Walter Hamscher <hamscher at mit-htvax>
Subject: EURISKO Seminar, Continued
[Forwarded from the MIT bboard by SASW@MIT-MC.]
TIME: 12 Noon
DATE: Friday, February 22
PLACE: 8th Floor Playroom
HOSTS: Michael Caine, Neil Singer, and Kenneth Pasch.
REFRESHMENTS: t
PLAUSIBLE POSITION GENERATION:
THE FURTHER ADVENTURES OF EURISKO, PART II
Blackstone Le Mot
This talk was to be the second in a series describing the application
of EURISKO, a discovery program, to non-traditional domains. In this
case the domain was position generation, in which the program is given
some knowledge of geometry, anatomy, and the first 37 pages of the
Kama Sutra. The talk is canceled, however, because the slides have not
yet been cleared with Dean McBay and the Ad Hoc Committee. Hence we
instead skip to the third and final talk in this series:
TARGET SELECTION:
THE LAST ADVENTURE OF EURISKO, PART III
Traditional strategic thinking, i.e. from Clausewitz to the present day,
emphasises the need to bring maximum destructive force to bear on the
enemy's armed forces and industrial centers to ensure a swift end to
hostilities. Present day weapons systems have been characterized as
"eggshells armed with hammers," suggesting that in the event of
hostilities, targets must be swiftly chosen, in time-frames requiring
automated response, to avoid the loss of precious megatonnage. In this
experiment we used EURISKO to choose targets in simulated nuclear exchanges,
with extremely exciting results.
------------------------------
Date: Mon, 18 Feb 85 19:42:32 est
From: Gary Cottrell <gary@rochester.arpa>
Subject: Seminar - Programming the User-Friendly Dog
SEMINAR
Saturday, 23 February 1985
55 Cottage St.
9:00 p.m.
Speaker
Garrison W. Cottrell
University of Cottage Street
Department of Dog Science
"Programming the User-Friendly Dog"
A current hot research topic is building user-friendly
interfaces to computer systems. One of the techniques of this
work is to design so-called "habitable subsets" of natural
language that in many cases allow the naive user to begin
productively using the system with little instruction. In this
work, we will show that these techniques, combined with results
from connectionist dog modelling, can be transferred to the ever
growing field of building user-friendly dogs. While the hardware
in this case is an example of a VRISC (Very Reduced Instruction
Set) computer, we will show that it is still possible to program
easily-learned high level commands.
Since it has been shown that in working with such machines,
the user has to do much of the computation of appropriate command
contexts (New Directions in Connectionist Dog Modelling, Cottrell
84), it is important to use English commands that make sense with
respect to the intended effect. For example, many previous
researchers have advocated the use of such commands as "Go on"
(How to Live with Three Dobermans, Kester 84) or "Go play" (Being
Mellow with Your Dog, Ose 73) to mean "Go lie down and quit
bothering me." The obvious mismatch here between the intent and
the usual meaning of "Go on" (i.e., continue) makes it difficult
for new users of the system to adapt to the command language. A
more ergonomically-designed command is "Scram." The command
matches the intent, and "go" is saved for more appropriate
contexts.
The reasons for the present sad state of affairs in most dog
programming systems can be traced to the use of outmoded command
languages and archaic beliefs about the capacities of the dog.
On the first point, many so called "experts" still advocate the
use of "heel" to mean "walk beside me." In this case, there is a
double mismatch: First with the hardware, which as everyone
knows, has no heel; and second with the semantics of the English
word "heel", which might better be used with respect to the male
dog's behavior towards female dogs. The New Age dog programmer
uses the much more natural command "Walk with Me." Addressing the
second point, many dog programmers believe that they have
accomplished much more than is possible with these crude
machines. It has long been known to those on the forefront of
this field (Larson, 84) that such baroque commands strings as:
"Now, JellyBean, you stay here, I have to go to a party and you
can't come. Be a good boy, JellyBean!" are actually interpreted
by the machine as: "blah JellyBean blah blah blah blah blah blah
blah blah blah blah blah blah blah blah blah blah blah
JellyBean," which is certainly not what the user intended. We
will have a demonstration system at the talk employing our
interface, including such useful commands as "Call the Elevator",
"Wag your tail", and "Eat that dog food."
------------------------------
Date: 22 Feb 1985 16:10 EST (Fri)
From: "Daniel S. Weld" <WELD%MIT-OZ@MIT-MC.ARPA>
Subject: Seminar - Motion Planning with Uncertainty (MIT)
[Forwarded from the MIT bboard by SASW@MIT-MC.]
On Motion Planning with Uncertainty
Michael Erdmann
Robots must successfully plan and execute tasks in the presence of
control and sensing uncertainty. Said differently, a robot must know
both how to get to a goal and how to recognize success once it has
gotten to the goal. I will present a backprojection algorithm that
computes regions from which motions along particular commanded
directions are guaranteed to successfully reach a goal. I will also
discuss the issue of goal recognizability and the power of the
backprojection approach in terms of the termination predicates
required to recognize success.
Tuesday, Feb. 26, 4 PM, 8th Floor Playroom
------------------------------
End of AIList Digest
********************