AIList-REQUEST@SRI-AI.ARPA (AIList Moderator Kenneth Laws) (09/20/85)
AIList Digest Friday, 20 Sep 1985 Volume 3 : Issue 125
Today's Topics:
Queries - FRL Sources & Parallel Rule Execution &
Information Retrieval & Lisp/Prolog for IBM4361,
AI Tools - Xerox 1185 & Prolog vs. Lisp,
Expert Systems - Intellectual Honesty and the SDI
----------------------------------------------------------------------
Date: Thu, 19 Sep 85 10:01:08 cdt
From: neves@wisc-ai.arpa (David Neves)
Subject: FRL sources
I remember a message a year ago about someone having the sources to FRL.
Unfortunately I don't remember who that person was. Could anyone who
has the sources send me mail? -Thanks, David Neves
Usenet: {allegra,heurikon,ihnp4,seismo,uwm-evax}!uwvax!neves
Arpanet: neves@uwvax
------------------------------
Date: Tue, 17 Sep 85 09:04:36 pdt
From: Thomas L. Zimmerman <zimmer%marlin@nosc.ARPA>
Subject: Parallel Rule Execution
The Naval Ocean Systems Center (NOSC) and Goodyear Aerospace are working
on a series of experiments to demonstrate the feasibiliy of running expert
systems on the Goodyear ASPRO parallel processor. A unique representation
for rules and data has been developed which on paper allows the ASPRO to
test 2,000,000 rules per second for satisfaction. This representation
scheme does put some limitations on the system involved - it appears to
require a very pure production system with no embedded control or functions in
the rules. So far a small (500 rule) system was written with this application
in mind and run on both a Symbolics and the ASPRO sucessfully. We would now
like to convert an existing sequential expert system for parallel execution
in order to determine the degree of speedup actually available and to discover
the limitations of converting a system not originally designed for this
application. Unfortunatly I am having trouble finding a system to convert for
this demonstration - which is why I am appealing to all of you. We need a
reasonably sized (200-700 rule) system, preferably military in character, that
meets the above limitations that we can attempt to run on our parallel inference
engine. This would be a no-cost way for someone to have their system speeded up
by an estimated three orders of magnitude. Any takers? If you're interested
please contact me:
Lee Zimmerman
Naval Ocean Systems Center
Code 421
San Diego CA 92152
(619) 225-6571 or zimmer@nosc
------------------------------
Date: 18 Sep 85 15:30:36 EDT
From: MARS@RED.RUTGERS.EDU
Subject: NLP for knowledge acquisition
Hi: I am interested in info about projects which use Natural Language
Processing Techniques to analyse scientific articles or abstracts
with the aim of deriving knowledge bases from them.
I am aware of a few projects in that field (UCLA, IBM
Heidelberg, Leiden University, Chemical Abstracts), but I
would appreciate any further pointers. Please
reply directly to me, and I will summarize to the net. Thanks.
Nicolaas J.I. Mars
[I have forwarded this message and the next to the information
retrieval list, IRList%VPI.CSNet@CSNet-Relay.ARPA. -- KIL]
------------------------------
Date: Wed, 18 Sep 85 09:08 EST
From: Ramesh Astik <rampan%northeastern.csnet@CSNET-RELAY.ARPA>
Subject: query - Information Retrieval
I was looking for some references in synthetic organic chemistry
on a Chemical Abstract Database system.The first hit was 10,000
followed by narrowing down to 400 ! The same search by an expert
bibliographer or search assistant gave me exactly 23 article list.
This is the difference in eye-balling and machine-search.
We had a lot of debate on the mental process of the human expert
doing the job and whether that can be mimicked on a computer or not.
Does any one know of any expert-system which can save us so
much paper work in searching and retrieving the information?
Our University has access to approximately 300 databases and
after the Automated search list is obtained,most of us do
equally long eye-balling!!
Kindly send any information to RAMPAN@NORTHEASTERN.
------------------------------
Date: Mon, 16 Sep 85 14:23:22 EDT
From: "Martin R. Lyons" <991@NJIT-EIES.MAILNET>
Subject: Info on Lisp/Prolog for IBM4361
Greetings all. Does anyone have any leads where I might obtain
a LISP or Prolog implementation that would run on an IBM 4361? I am
looking for public domain or reasonably priced packages; and wouldn't
you know it, time is of the essence. If anyone has any leads please
contact me at the address below.
I will summarize and post here if there is sufficient interest.
As always, thanks in advance!
MAILNET: Marty@NJIT-EIES.Mailnet
ARPA: Marty%NJIT-EIES.Mailnet@MIT-MULTICS.ARPA
or @MIT-MULTICS.ARPA:Marty@NJIT-EIES.Mailnet
USPS: Marty Lyons, CCCC/EIES @ New Jersey Institute of Technology,
323 High St., Newark, NJ 07102 USA (201) 596-2932
"You're in the fast lane....so go fast."
------------------------------
Date: 16 Sep 85 15:28 PDT
From: Fischer.pa@Xerox.ARPA
Subject: Correction: Xerox 1185 uses 80186 not 8086
Being as close as I am to the subject I'm embarassed to say (but
overjoyed to know) that I was wrong and that the 1185 and 1186
workstations use the Intel 80186 processor chip both in the IO processor
and on the IBM PC emulator boards. Their (micro programmable) main
processor is built with bitslice devices.
But note that I never claimed to be a faultless spokes-thing for Xerox
or Xerox AI Systems, just another random attendee of IJCAI.
Sorry folks,
(ron)
------------------------------
Date: Tue, 17 Sep 1985 06:28 EDT
From: Hewitt@MIT-MC.ARPA
Subject: Prolog and Lisp
I would like to respond to the message in AILIST from Fanya Montalvo
who wrote:
From: Hewitt
However, it
is not possible to make a commercially viable Common Lisp
implementation on Prolog. This means that any good software written
for a stand alone Prolog system will soon appear on the Lisp Systems
but NOT vice versa. Therefore the stand alone Prolog systems will
always have impoverished software libraries by comparison with the
Common Lisp systems and will not be commercially viable in the long
run.
From: Montalvo
This type of argument strikes one as historical accident not as anything
fundamental. And by that accident I mean vagaries of the market place.
Could you say more about how it's fundamental, or do you agree that it's
historical accident?
This is a very good question! However, in this case I believe that
the commercial marketplace is reflecting some very real deficiencies
in Prolog and its underlying conceptual basis of LOGIC as a
PROGRAMMING language. As far as I can tell no one in the Prolog
community believes that they will EVER be able to construct a
commercially viable Common Lisp on Prolog. That indicates that there
are some critical limitations of Prolog that are not just historical
accidents. I believe that the deficiencies of Prolog cannot be
repaired within the framework of LOGIC as a PROGRAMMING language.
------------------------------
Date: Wed, 18 Sep 85 18:05 EDT
From: Hewitt@MIT-MC.ARPA
Subject: Lisp vs. Prolog vs. ?
I agree with Wayne McGuire who wrote as follows to this list:
... one might speculate that perhaps the bulk of AI code
will be written neither in Lisp nor Prolog (not even an enhanced
Prolog which can elegantly manipulate and coordinate in the same
conceptual space multiple worlds, logics, and beliefs), but a
higher-level language, perhaps using Lisp and/or Prolog as a base.
However, I would like to point out that Prolog is not a suitable base
to implement the kind of higher-level languages that Wayne envisions
since it does not provide the appropriate primitives. Implementing
higher-level Artificial Intelligence languages requires efficient data
structure and control primitives that are not part of Prolog. I agree
that Common Lisp is too low level to use very much to directly
implement applications. However, Lisp has historically been an
excellent implementation language for implementing higher-level
Artificial Intelligence languages. Indeed some of the best Prolog
systems and other logic systems such as FOL (Wheyhrauch et. al.), MRS
(Genesereth et. al), the Pure Lisp Theorem Prover (Boyer and Moore),
and LogLisp (Robinson et. al.) are implemented on Lisp.
------------------------------
Date: 18 Sep 85 15:48 EDT
From: WAnderson.wbst@Xerox.ARPA
Subject: Intellectual honesty and the SDI
At the recent IJCAI at UCLA I picked up a couple of papers at the GE
exhibit booth. One of these, entitled "A Tutorial on Expert Systems
for Battlefield Applications," (delivered at a meeting of the Armed
Forces Communications and Electronics Association last May) states that
"AI systems that incorporate human expertise may be the only way" to
fill the gap between availability of people and complexity of military
hardware. In defense of this strategy the author states:
- In contrast with humans, AI systems are good at handling the myriad
details of complex situations, such as often occur in military settings.
- In contrast with other computational approaches that are more formal
and algorithmic, AI systems are more robust: they are designed to deal
with problems exhibiting uncertainty, ambiguity, and inaccuracy.
I find it appalling (and frightening) that statements like this can be
presented in a technical paper to military personnel. The author
(according to the references) has contributed widely to the AI field at
many conferences. It's simply ludicrous to state that current AI systems
are better in battlefield situations than humans. What was the last AI
system that could drive a tank, carry on a conversation, and fix a
broken radio whilst under enemy fire? The second comment is equally
misleading. To contrast "formal and algorithmic" with "robust" seems
to imply that algorithms and formal procedures are inherently not
robust. On what is this claim based? (There is no reference attached
to either statement.) It sounds like a recipe for unreliable software
to me.
How can someone write this stuff? I know, to make money. But if this
is the kind of information that is presented to the military, and upon
which they make decisions, then how can we expect any kind of fair
assessment of the possible projects in the Strategic Computing (and
Defense) Initiatives? How can this kind of misinformation be rebutted?
Bill Anderson
P.S. The full reference is available on request.
------------------------------
End of AIList Digest
********************