Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 096
AIList Digest Tuesday, 22 Apr 1986 Volume 4 : Issue 96
Today's Topics:
Queries - Technical Report Sources & FRL and HEARSAY-III &
Lisp Machine Information & AI Conference in Long Beach &
National Youth Science Camp Alums & Battleware,
Projects - EDISON Project & Non-DoD Funding,
Techniques - String Reduction,
Publications - Journal Prices,
Applications - Compuscan Page Reader
----------------------------------------------------------------------
Date: Fri 18 Apr 86 17:17:43-PST
From: Daniel Davison <DAVISON@SUMEX-AIM.ARPA>
Subject: How do I get technical reports?
There were several technical reports mentioned in a recent AIlist that I'd
like to get...but I don't know how. Would some kind soul send me a note
about how to get tech reports from (1) LSU and (2) CMU?
Thanks,
dan davison
davison@sumex-aim.arpa
------------------------------
Date: 24 Apr 86 21:35:30 GMT
From: ihnp4!houxm!whuxl!whuxlm!akgua!gatech!seismo!mcvax!euroies!rreilly
@ucbvax.berkeley.edu
Subject: FRL and HEARSAY-III
I have two queries:
(1) Does anybody have any information on public domain
implementations of Goldstein's FRL or any frame system in
Lisp?
(2) I believe that CMU make available an empty version of HEARSAY
(HEARSAY-III I think). Does aybody have any details on this?
Has anybody used the system?
Thanks in advance.
--
...mcvax!euroies!rreilly (Ronan Reilly)
Educational Research Centre, St Patrick's College
Dublin 9, Ireland.
------------------------------
Date: Mon, 21 Apr 1986 12:23:33 EST
From: WALLFESH%UCONNVM.BITNET@WISCVM.WISC.EDU
Subject: Lisp Machine information sought
Can anyone suggest any papers on Lisp machines, particularly
those which stress their architectural aspects? I'm attempting
to write a paper on Lisp machines for my computer architecture
class, but I cannot seem to find many references.
Thanks,
Sande Wallfesh (wallfesh@uconnvm.bitnet, wallfesh@carcvax.csnet)
CS Dept. Box U-157
University of Connecticut
Storrs, CT 06268
------------------------------
Date: Fri, 18 Apr 86 14:14:17 cst
From: Girish Kumthekar <kumthek%lsu.csnet@CSNET-RELAY.ARPA>
Subject: AI Conference in Long Beach
I am trying to find information about an AI conference in Long Beach, CA to
be held by the end of this month (April).
I looked in recent issues of AI magazine but could not get info.
If you have any info as to exact dates,
whom to contact, tel # or address etc, I would appreciate if you could mail
it to me at
kumthek%lsu@csnet-relay.csnet
Thanks in advance
Girish kumthekar (504)-388-1495
------------------------------
Date: Sun 20 Apr 86 11:23:59-PST
From: Lee Altenberg <ALTENBERG@SUMEX-AIM.ARPA>
Subject: NATIONAL YOUTH SCIENCE CAMP ALUMS
I am an alum of the National Youth Science
Camp (California 1975), which was a great boost for me going into
science, and I was wondering how many other NYSC alums are involved in
AI . NYSC Alums please send me your name , and the state and year you
represented, and also what you're doing now if you want to. Thanks.
-Lee Altenberg
------------------------------
Date: Fri, 18 Apr 86 11:42 EST
From: JOHNSON%northeastern.csnet@CSNET-RELAY.ARPA
Subject: Rampant misunderstanding, speculation and questions.
I've been reading the April CACM (my first issue). One of the
sections seems to be a debate over battle management software. It seems
that nobody is sure whether or not battleware would do the "right" (if
there is a right) thing in a real battle. Also, no one is sure if it
ever can be proven one way or the other.
I'm not a military historian. What little I do know tells me that
one doesn't necessarily win a battle with logic or one set of rules.
History seems to say that you start losing wars if the other side
doesn't play by the rules or invents their own.
I've been seeing shorts in the ailist of late dealing with inexact
logic. I don't claim to understand it. I'm not at all certain,
however, that I'd want a piece of ai battleware defending me even if it
did have some kind of rudimentary understanding of non-binary logic. In an
impossible situation where the other side doesn't play by the rules,
battleware might decide to give up or pull the doomsday switch.
I guess I have several questions here. How inexact is inexact? Can
battleware but made to recognize that new rules exist and adapt to them
in real-time? Can battleware be made to fully understand constraints?
If my cat has fleas and fire kills fleas, burning the cat will eliminate
the fleas. Problems are usually easier to solve without constraints.
I'd rather battleware (or any ai program) didn't think like this. I
know a little about how rules can be made to work in Prolog. I'm not
sure how one would go about defining rules for a battle. Can there really
be inexact (fuzzy) rules? Any references on this?
A side issue is this, can battleware be made to run in real-time
at all? One of the ideas of having it is because humans can't
assimilate all necessary information in time to act on it in a highly
electronic war. I don't think this is just an issue of having 10 cray's
doing the job.
I don't know. Maybe 20,000 toasters running in parallel would be
just as good as two or three crays running battleware. Maybe I just
don't understand the problem. (I'm new to ai obviously.) Can we please
define or describe some things in words of one syllable or less? (yes I
know graph is one syllable.) Besides defining inexact, can anyone point
me to a GOOD beginner something or other on ai. I've seen several and
they're all not very good yet (or I might be missing the point).
Chris Johnson
Northeastern University
johnson@northeastern.csnet
------------------------------
Date: Fri, 18 Apr 86 11:34:41 PST
From: Scott Turner <srt@LOCUS.UCLA.EDU>
Subject: Re: EDISON Project
Before I get innudated with requests for the Edison report, all UCLA
Technical Reports can be ordered through:
Brenda Ramsey
UCLA Dept. of Computer Science
3713 Boelter Hall
801 Hilgard Ave.
Los Angeles, CA 90024
ramsey@ucla-cs.arpa
(213) 825-2778
-- Scott
------------------------------
Date: Fri 18 Apr 86 17:18:30-PST
From: Daniel Davison <DAVISON@SUMEX-AIM.ARPA>
Subject: Summary of non-DoD funding for AI
I recently asked about non-DOD sources of funding for AI projects. I received
three replies. Two cited NSF and one said NIH (but not which institute[s])
fund AI work. Of course, NIH funds this facility (SUMEX); apparently smaller
grants are funded also.
dan davison
davison@sumex-aim.arpa
------------------------------
Date: 18 Apr 86 16:59:06 GMT
From: ihnp4!stolaf!mmm!umn-cs!amit@ucbvax.berkeley.edu (Neta Amit)
Subject: Re: String reduction
In article <1031@eagle.ukc.ac.uk> sjl@ukc.ac.uk (S.J.Leviseur) writes:
>Does anybody have any references to articles on string reduction
>as a reduction technique for applicative languages (or anything
>else)? They seem to be almost impossible to find! Anything welcome.
String reduction as a model of computation was suggested by
A.A. Markov, in his 1954(?) paper, and is proved to be equivalent in
power to the other two general models of computation (Turing machine and
the Lambda Calculus).
Markov algorithm consists of an input string S and a program P, which
is a sequence of BNF-like productions LHS --> RHS. The evaluator scans
P top to bottom (that's sequencing), looking for a match between a
substring of S and the LHS of a production (that's conditionals). When
a match is found, the RHS replaces the matching substring in S, and
the scanning is restarted from the top of P (that's looping). Let's
not consider termination and error conditions. As stated here, this
isn't a purely applicative model, but there is no inherent reason why
the new S couldn't in fact be new!
Michael Barnett of Brooklyn College, CUNY, (a Chemist turned Computer
Scientist) has recently suggested (See Sigplan Notices in the last 12
months) that it may be possible to synthesize molecules that will do
string substitution (a biological computer) and that this might be a
good model to describe the functionality of the human brain.
If I understand correctly, you are looking for an applicative model,
in which functions cause string-substitution instead of returning
values. Notice that this is the mechanism used by parametrized macro
expansion (so you can easily simulate an applicative string reduction
machine in Pure Lisp, using macros alone.)
A guy named Karl Fant, from Honeywell Research (in Minneapolis), has
been developing an applicative string-reduction model, but I don't
think he has published in a publicly available journal.
Anyway, I would be interested in expanding this discussion.
Cheers,
--Neta CSNET: amit@umn-cs
ARPA: amit%umn-cs@csnet-relay.ARPA
UUCP: ...ihnp4!umn-cs!amit
------------------------------
Date: 19 Apr 86 15:35:22 GMT
From: ihnp4!stolaf!mmm!srcsip!meier@ucbvax.berkeley.edu (Christopher Meier)
Subject: Re: String reduction
In article <994@umn-cs.UUCP> amit@umn-cs.UUCP (Neta Amit) writes:
>A guy named Karl Fant, from Honeywell Research (in Minneapolis), has
>been developing an applicative string-reduction model, but I don't
>think he has published in a publicly available journal.
>
Karl can be reached at ihnp4!srcsip!fant (or through any path I can...).
--
Christopher Meier MN65-2300 {osu-eddie,okstate,bthpyd}\
S&RC Signal and Image Processing {ihnp4,philabs,,gnutemp}\
3660 Technology Drive (612) {hyper,umn-cs,mmm,meccts}!srcsip!meier
Mpls, MN 55413 782-7191
------------------------------
Date: Mon, 21 Apr 86 07:15:35 -0500
From: sriram@ATHENA.MIT.EDU
Subject: Journal Prices
In a recent message on the AI bboard, some of you raised concerns
about the subscription price for the AI IN ENGG. JOURNAL. As a
co-editor of the journal, I forwarded the message to Mr. Lance
Sucharov, the publisher director of the journal. His reply follows.
Sriram
Dear Friends:
I have just received a copy of the message "Journal prices hit the
moon!" on the AI list bulletin board and I feel this needs an answer
and is an excellent opportunity to put the publishing record straight.
The journal referred to is our new launch the "INTERNATIONAL JOURNAL
FOR AI IN ENGINEERING", which has aroused a good deal of interest
internationally and in this one case, controversy over the price!
First of all, may I say I sympathise with the writer: in wishing to
keep prices down as much as possible and thereby giving the widest
audience to published information. However, although the writer
mentions the authors, reviewers and editorial board members who give
their services freely, he makes no mention of others who do not:
typesetters, printers, agents, editorial staff, designers, editors,
and so on.
How significant are these costs? Significant enough for some journals
to close. Many others of important academic interest, financially
only limp along and as publishers, would be better off putting their
money in a bank. At the same time some publishers are extremely
profitable: a well-known name and powerful marketing arm all help. But
it is by having some profitable journals that more marginal ones can
be popped up, and new launches can be afforded, which can take several
years to pay their way. Furthermore, for overseas publishers the US
market can be expensive because of postage, uncertainty in dollar
movements and banking costs. Other publishres have pitched their
prices higher than ours, and I can mention the "IMA JOURNAL OF NUMERAL
ANALYSIS" ($172), "COMPUTER SYSTEMS SCIENCE AND ENGINEERING" ($166),
"IMAGE AND VISION COMPUTING" ($171), and there are many more.
Finally, I am delighted that authors and the editorial team do provide
a service just as we provide a service in disseminating their work
world-wide.
Sincerely yours,
Lance Sucharov
------------------------------
Date: 18 Apr 1986 13:16:39-EST
From: kushnier@NADC
Subject: Compuscan
COMPUSCAN
I had an opportunity to see a demonstration of the CompuScan Model 230
Page reader. This is a desk top device which looks like a small copier,
and using optical character recognition, can read a page of text and
send it as an ASCII file to a waiting IBM-PC or other micro over an RS-232
port. The unit sells for about $6K.
I had provided several different font sets on several qualities of paper.
Also, I used originals and text that had been Xeroxed several times.
As for the results...
If I put on my R&D AI hat, then the results were both exciting and thought
provoking. I was amazed at the number of characters that the machine was
able to successfully recognize- especially for a machine in that price
category. Although it usually took longer than the advertised 30 seconds
per page, it was tolerable.
The results were inconsistent. If you put the same page in twice, the results,
both errors and correct characters, would come out different each time. Now,
this is not to say that the machine was "screwing up". It WAS, in each case
following a set of rules based on what it perceived to be a specific
character. It was going through a set of probabilities and percentages
and based on the result, printed a particular character. Being a
deterministic programmer, this at first rubbed me the wrong way. After
all, we are dealing with a computer here..No matter how many times you
put in 2+3, it should always equal 5. Not so with AI type solutions.
Although this inconsistency should be minimized in the design, the AI
programmer must recognize the possibility of it occurrence.
If I put on my Office Manager's hat, then I would say that COMPUSCAN is
not quite ready to come into everyday service. It made too many mistakes
to provide efficient page-to-text translation. This was especially true when
the quality of the documents and the font type varied. One page had been
slightly cocked when Xeroxed. This caused havoc to the optical recognition
software.
Compuscan is promising a new generation Model 240, to be out shortly. I
am interested in seeing what improvements are made.
For INFO write to: Compuscan Inc.
81 Two Bridges Rd./Bldg.2
Fairfield, N.J. 07006
TEL: (201) 575-0500
This review is a personal opinion and does not reflect any official view of the
government or any one else in the world. - Ron Kushnier
Ron Kushnier
kushnier@nadc.arpa
------------------------------
End of AIList Digest
********************
