Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 168
AIList Digest Friday, 18 Jul 1986 Volume 4 : Issue 168
Today's Topics:
Queries - PC Expert-System Shell Demos &
Garbage Collection Side Effects & Online Almanacs,
Policy - Signing with Net Addresses,
Literature - LISP Texts & Natural Language,
AI Tools - Parallel FCP,
Review - July Spang Robinson Report
----------------------------------------------------------------------
Date: 14 Jul 86 06:56 PDT
From: A. Winsor Brown, Douglas Aircraft ET&IS <AWB.MDC@OFFICE-1.ARPA>
Subject: PC Expert-System Shell Demos
I am aware of the following low cost or no cost PC expert system shell demo
and/or full packages. I am going to use all of them in an internal course on
PC based shells/tools, and want each of the students to retain a copy (thus the
concern for low priced demonstrators or full systems). Since the course is to
address evaluation criteria, actually seeing the tool in use is important.
If you know of others would you please let me know, and include a source phone
number (and address if the provider was not listed in the 15 June "AI Software
for MS-DOS (Long List)"). I will summarize the responses and re-post them
here.
M.1 with a date of Oct'85 in the "Demonstration Materials" manual; from
Teknowledge 415/327-6640. Single disk plus report sent on request; freely
copiable for demos. Loading and saving knowledge base disabled in demo; only
five knowledge base additions allowed.
EXSYS Version 3 from EXSYS, Inc (Albuquerque, NM) 505/836-6676. Copying
and distribution encouraged; first copy costs $15. Learning tutorial
included with all demo's; full manual in softcopy form in the older 3
diskette version; hard copy form of manual with the newer 2 disk set which
also allows saving a 25 rule expert system.
Guru initial release (1.00c) from MDBS 317/463-2581. 4 disk set; sent on
request (not copy protected but further distribution not proscribed:
Copyright protection claimed for diskettes and 50 page "Demonstration
Instructions"). On-line help documentation is all that is provided.
Definitely need hard disk: 1.2Meg needed just for demo! Can develop small
(10 rule) single rule-set expert system with Demo; some other restrictions
apply to other parts.
1st Class version 3.0 (3/86) from Programs in Motion 617/653-5093. Demo
disk costs $20; re-distribution not currently desired/allowed. Manual not
included on diskette; some technical details provided on-line.
Personal Consultant version 1.00 from TI 800/527-3500. Available from
TI, at no charge; copy protected. 43 page "Demonstration Guide" and 22 page
"Technical Report" (on-line help; no documentation per se). Needs full
512K; can run off single floppy; crippled so user cannot save developed
system.
ESIE version 1.1 from Lightwave Consultants 813/988-5033. Shareware
(registration fee is $75). 25 page manual included on diskette.
Knowledge Delivery System version of 8/85 from KDS Corporation
312/251-2621. Available from KDS for $25; allowed to be reproduced and
distribution verbally proscribed. No manual; not clear about on-line help.
Development example limited to 20 cases (examples) from the normal 4096;
also has some text size restrictions.
Expert System version ??? from PPE 301/977-1489. A public domain tool;
available for $20 from PPE too. Manual situation not clear ("program is
self documenting" comment from another knowledgeable source). Source code
is included!
Thank you. --Winsor
------------------------------
Date: Tue, 15 Jul 86 23:46:59 CDT
From: David Chase <rbbb@rice.edu>
Subject: Query on compilers, optimization, and garbage collection
I am looking for references on interactions (good and bad, intended and
unintended) between garbage collectors and compilers that (attempt to) do
optimizations. For example, if you know of a good optimization that
reduces the amount of garbage produced, tell me about it. If you know of
an ugly surprise that someone received when they tried to optimize code in
a garbage-collected system, tell me about that.
I realize that this isn't exactly AI, but I think people reading this list
might have some pointers (to other lists, if nothing else).
What I already have (no references for ugly surprises):
"Optimization of Very High Level Languages-I: Value Transmission and
its Corollaries"
Schwartz, in Computer Languages, volume 1, pp 161-194 (1975)
(copy optimizations, heap->stack allocation conversions)
"Experience with the SETL Optimizer"
Freudenberger, Schwartz and Sharir, in TOPLAS 5:1 (January 1983)
(copy optimizations)
"Binding Time Optimization in Programming Languages: Some Thoughts
Toward the Design of an Ideal Language"
Muchnick and Jones, in POPL 3, 1976
(heap->stack allocation conversions)
"Shifting Garbage Collection Overhead to Compile Time"
Barth, in CACM 20:7 (July 1977)
(reference counting at compile time)
"RABBIT: A Compiler for SCHEME"
Steele, 1978
(heap->stack allocation conversions for activation records)
"Fast Arithmetic in MacLISP"
Steele, in 1977 Macsyma Users' Conference
(heap->stack allocation conversions for numbers)
"An Optimizing Compiler for Lexically Scoped Lisp"
Brooks, Gabriel and Steele, in Compiler Construction 1982
(heap->stack allocation conversions for numbers)
"A scheme of storage allocation and garbage collection for ALGOL 68"
Branquart and Levi, in Algol 68 Implementation (North-Holland, 1971)
(compiled marking routines)
"Methods of garbage collection for ALGOL 68"
Wodon, in Algol 68 Implementation (North-Holland, 1971)
(compiled marking routines)
David Chase
------------------------------
Date: Wed, 16 Jul 86 16:28 EST
From: LEWIS%cs.umass.edu@CSNET-RELAY.ARPA
Subject: almanacs and magnitudes
I would appreciate any information people could give me on the availability
of online almanacs or similar large bodies of broadranging statistical data.
Public domain or cheap would be preferable, of course. I also would be
interested in hearing about any work that has been done either on AI programs
for heuristically estimating the information one finds in almanacs, or on
psychological research on human order of magnitude estimates. So far the
only place I've seen this subject discussed are the entertaining diatribes
on "number numbness" by Douglas Hofstadter in Metamagical Themas, and by
Jon Bentley in a recent issue of CACM.
Please send replies to me; if there is sufficient interest I will summarize
for the digest.
Thanks, David D. Lewis
Univ. of Massachusetts, Amherst
"well, I used to think it was LEWIS@UMASS-CS
and lately it's been LEWIS%cs.umass.edu@CSNET-RELAY.ARPA
but maybe it's longer now"
------------------------------
Date: Mon, 14 Jul 86 16:21:04 cdt
From: Girish Kumthekar <kumthek%lsu.csnet@CSNET-RELAY.ARPA>
Subject: Replying to AIList messages
I have been reading messages and find them interesting.
However I find that most of the times, the direct address where the
reply can be sent is not given.
It is typically at the top of the message, and is probably mixed with
other details.
This forces you to type cntl-z to stop the viewing and come back and note
the address at the top.
So would you all please give your addresses at the end of messages.
(Note that at the top of message, one does'nt know if this message is
going to turn out interesting or not ! ).
My address is kumthek%lsu@csnet-relay.csnet
Thanks in advance
Girish Kumthekar (504)-343-5334
------------------------------
Date: Fri, 11 Jul 86 09:29:42 EDT
From: "William J. Rapaport" <rapaport%buffalo.csnet@CSNET-RELAY.ARPA>
Subject: LISP texts (followup article)
Another good LISP text is the new one by my colleague Stuart C. Shapiro:
LISP: An Interactive Approach (Computer Science Press). It is
dialect-independent and intended for self-study. We have used it (in
both manuscript and published versions) for a number of years at SUNY
Buffalo, with great success.
------------------------------
Date: 14 JUL 86 16:34-EST
From: PJURKAT%SITVXA.BITNET@WISCVM.ARPA
Subject: REFERENCES ON NATURAL LANGUAGE
I'm a little slow reading my mail. This is a response to a query in
Vol 4 Issue 151 by Gene Guglielmo (sefai@nwc-143b asking for
references concerning the representation of natural language on a computer
system. Please pass the following on to him:
Parisi and Antonucci
Essentials of Grammar
It presents a represenation of sentences in functional form, that is,
predicate(arg1, arg2, ... )
taking into acoount a goodly amount of semantics. I have found it valuable,
especially for the analysis of belief systems.
Cheers - Peter Jurkat (pjurkat@sitvxa.bitnet)
------------------------------
Date: Fri, 11 Jul 86 09:35:52 -0200
From: Steve Taylor <steve%wisdom.bitnet@WISCVM.ARPA>
Subject: Parallel FCP
We are pleased to announce the availability of a parallel Flat
Concurrent Prolog (FCP) [1,2] interpreter for the Intel iPSC
Hypercube. The interpreter may be used for initial experiments with
parallel logic programming; it includes most of the kernel predicates
available in the Logix system.
FCP programs may be developed on a uniprocessor under the
Logix system, which is announced seperately [3]; this environment operates
on the VAX, SUN or Intel 310 systems. Recompilation allows the
resulting program to execute on the Intel iPSC hypercube. Simple
techniques have been developed to map processes and code to the
physical machine [4]. These techniques allow multiple virtual
machines to execute concurrently; multiple applications may execute
concurrently on a given virtual machine.
PLEASE NOTE: The interpreter is an experimental system which has only
recently been completed; it is being made available on an informal
basis to encourage members of the community to experiment with the
language.
The handling fee for a non-commercial license to the
Parallel FCP Interpreter and the Logix system for the 310 is
$250 U.S. To obtain a license form and/or a copy of the Logix user
manual write to:
Steve Taylor
Department of Computer Science
The Weizmann Institute of Science
Rehovot 76100, Israel
To obtain an electronic copy of the license write to:
CSnet, Bitnet: steve@wisdom
ARPAnet: steve%wisdom.bitnet@wiscvm.arpa
Sincerely,
Steve Taylor
References
[1] C. Mierowsky, S. Taylor, E. Shapiro, J. Levy and M. Safra, "The
Design and Implementation of Flat Concurrent Prolog", Weizmann
Institute Technical Report CS85-09, 1986.
[2] A. Houri and E. Shapiro, "A sequential abstract machine for Flat
Concurrent Prolog", Weizmann Institute Technical Report CS86-20,
1986.
[3] W. Silverman, M. Hirsch, A. Houri, and E. Shapiro, "The Logix
system user manual, Version 1.21", Weizmann Institute Technical
Report CS86-21.
[4] S. Taylor, E. Av-Ron and E. Shapiro, "A Layered Method for
Process and Code Mapping", Weizmann Institute Technical Report
CS86-17.
------------------------------
Date: WED, 20 apr 86 17:02:23 CDT
From: E1AR0002%SMUVM1.BITNET@WISCVM.ARPA
Subject: July Spang Robinson Report summary
Summary of Spang Robinson Report July 1986, Volume 2, No. 7
(Emphasis on AI and Parallel Procesing)
Advanced Decision Systems is using the Butterfly for AI development.
They are developing a SCHEME using message passing for the Butterfly. They
are developing an expert system to balance work loads and manage faults.
The system works well on about 10 nodes but past these points, the system
performance does not continue to improve as processors are added. It also takes
20 minutes to reboot the machines.
Oak Ridge National Laboratories has been using the NCUBE for machine
vision research.
NASA is using a FLEX/32 parallel machine to develop an expert system
shell and an expert system to predict sun spot activity. CLIPS will
run on the FLEX/32 and is an OPS5-like system written in C. In the
sun spot sytem, the expert system part of the application will run on
the Symbolics with the math part running on the FLEX using parallelism.
LUCID is developing an implementation of parallel LISP under subcontract
to Stanford. The work is starting on a newly purchased Alliant Computer
Systems.
Cray Research has some proprietary AI projects in its Applications
department. ELXSI is looking for a client who needs AI on
mainframe class of machine. It also is looking for a vendor to port
AI language to ELXSI. Encore Computers and Masscop has active programs to
produce AI languages.
The Kemp-Carraway Heart institute is doing image analysis of echo
cardiograms using fuzzy logic. It has developed a FLOPS product in
which rules can fire in parallel and eliminates the need for "truth
maintenance" when rules do not have to be executed sequentially. The system
uses fuzzy logic with an OPS-5 syntax.
One Forth researcher claims to have designed a 1 million logical inference
per second expert system on the Novix NC4000 Forth engine (a $150.00 chip).
__________________________________________________________________________
News section
Carnegie Group has signed an agreement with Hewlett-Packard to offer
its Knowledge Craft expert system shell on the HP 9000 Model 320.
Lisp Machines has made 18 changes to their machine to improve reliability.
They will have an AVP processor that will be twice the speed of their
current processor. They are also working on a LISP chip and on improvement
of the development environment for non-LISP machines.
Intellicorp has doubled its direct sales force and has established a VAR
relationship with AMOCO corporation.
TI has sold 1000 Explorer work stations of which 200 are in universities.
A key reason for Burroughs recent merger with Sperry is Sperry's AI activity.
------------------------------
End of AIList Digest
********************
