LAWS@SRI-AI.ARPA (05/17/85)
From: AIList Moderator Kenneth Laws <AIList-REQUEST@SRI-AI>
AIList Digest Thursday, 16 May 1985 Volume 3 : Issue 64
Today's Topics:
Queries - PROLOG on INTERLISP/LOOPS & Lisp Machines &
Microcomputer Lisps & Statecharts,
Machine Translation - Update,
Expert Systems - Prospector on a PC,
Seminars - Diagnosing Multiple Faults (SU) &
Fixpoint Extensions of First-Order Logic (CMU) &
A New "Turing" Thesis (CMU),
Course - Model Theory (CMU)
----------------------------------------------------------------------
Date: 13 May 85 18:06:35 EDT
From: Louis Steinberg <STEINBERG@RUTGERS.ARPA>
Subject: PROLOG on INTERLISP/LOOPS
Can anyone point me to an implementation of PROLOG that will run on
a Xerox Lisp machine, i.e. is implemented in Interlisp-D or Interlisp-D
with LOOPS? I know of Ken Kahn's version but that unfortunately does
not use standard LOOPS. This is for an educational environment so
efficiency is not essential. Also welcome would be advice on porting
some other, existing version.
Lou Steinberg
STEINBERG @ RUTGERS
------------------------------
Date: Tue, 14 May 85 15:32:55 pdt
From: Curtis L. Goodhart <goodhart%cod@Nosc>
Subject: Lisp Machines
Anybody have any pointers to some good references for explaining the
architectural characteristics of a lisp machine, ie why do you need a
specific kind of machine, as opposed to a conventional computer, to run
lisp?
Thanks,
Curt Goodhart (goodhart@nosc ;on the arpanet)
------------------------------
Date: Wed, 15 May 85 10:23 EST
From: John N Frampton <frampton%northeastern.csnet@csnet-relay.arpa>
Subject: Microcomputer Lisps comparable to GCLISP
I am writing a review of GCLISP for a computer magazine. I have used
MULISP and IQLISP on an IBM PC previously but just found out yesterday
that there are several new Lisp implementations for the IBM PC which
are competitive with GCLISP. I would appreciate very much getting a short
description of these products to include in the GCLISP review. I would
have preferred to do a comparison, but it's too late for that. At least
I can say what's out there.
I'd like a short description (better shorter and sooner than longer and later)
- particularly touching on the extent to which they implement Common Lisp
(if they do) and if they have a compiler.
Reply directly to me (and the board if you want).
Thanks, John Frampton
------------------------------
Date: Monday, 13 May 1985 15:02-EDT
From: rh@Mitre-Bedford
Subject: STATECHARTS
Dave Harel was to have given a talk "Statecharts: A Visual
Approach to Complex Systems" last week at MIT. I'd be interested
in anyone's impressions who attended or otherwise knows about this
work. Are there any available references?
Thanks,
Rich Hilliard
------------------------------
Date: Tue 14 May 85 11:08:50-MDT
From: Pete Tinker <tinker@UTAH-20.ARPA>
Subject: Machine Translation Update
A friend of mine who works as a consultant for Weidner informs me of some
corrections to the information posted by Petitpierre on Machine Translation.
Only the address for ALPS was given, but the message could be construed
to also be the address of their competitor, Weidner. Weidner is no longer
in Provo, Utah, and is no longer called Weidner. The correct name and address
are
WCC
Suite 300
40 Skokie Boulevard
Northbrook, Illinois 60062
(312) 564-8122
Also, Siemens has an active MT group in Boca Raton, Florida; LOGOS
does MT in Boston (Petitpierre mentioned only their European branches).
ISSCO is also in Sorrento Valley, California, but I don't know if they
work in MT there.
-Pete Tinker
------------------------------
Date: Tue 14 May 85 10:57:37-PDT
From: HART@SRI-AI.ARPA
Subject: Prospector on a PC
Ken,
I saw a copy of the "Prospector on a PC" query, and thought your
AIList readers might be interested in the following clarifications and update.
First, the role of Prospector in the famous molybdenum strike:
Unfortunately for journalists, an accurate account of the role played by
Prospector is not simply stated, and a simple account is not likely to be
accurate. Prompted by the confusion surrounding these events (notwithstanding
a research report that appeared in Science in 1982), Dick Duda, Rene' Reboh
and I submitted a rather lengthy letter to the AI Journal that lays out the
details quite carefully. Since the letter should be appearing shortly, and
in recognition of my own admonition about simple statements, I will say here
only that the evaluation of the moly prospect was intended purely as a
scientific experiment and that Prospector's predictions were confirmed to
a high degree by independent means.
Second, Prospector's origins and descendants: My first notes on
the subject date from early 1974, and were motivated by the early success of
MYCIN. The project began in earnest in 1975, with Dick Duda, Nils Nilsson,
Georgia Sutherland and a geologist named Alan Campbell as the earliest
contributors. Rene Reboh joined the project shortly thereafter, and John
Gaschnig a couple of years after that. Since the earliest implementation
there have been many other contributors and any number of implementations
on other machines. Some of the earliest small-machine systems were done
by Rene Reboh (on an Apple II) and by John Reiter on a PDP-11. These and
other ports generally are not as powerful as the original PDP-10 version,
but have been useful nonetheless. The USGS has done a port to the IBM/PC,
and Alan Campbell has done an independent port to the IBM/PC and is
marketing his as "The Deciding Factor." Coincidentally, a user's review
of The Deciding Factor appeared in last Sunday's (May 12) San Jose Mercury
by Ray Levitt, a Civil Engineering professor at Stanford. (He liked it,
by the way.) I am told that some mining companies are interested in Campbell's
version; however, any discussion of use by commercial companies must
begin with the recognition that the North American mining industry has been
in a severely depressed state for quite some years.
The US Geological Service has an active program underway to extend
and use Prospector. They have the PC version mentioned above, as well as
a Xerox 1108 version. There are something like 30 - 35 models that have
been developed (a "model" is a major module of the knowledge base). The
Survey has used Prospector to evaluate mineral potential in Alaska and
in New England, and has stated (in writing) that the results obtained were
in their view superior to what would have been otherwise achievable.
(Incidentally, the USGS has a charter to assess resource potential over
large tracts of land rather than "to discover a mine".) The Survey also
has some 8-10 models under active development for the PC version, and
is planning to use the 1108 version for geographically-organized data.
Finally, I am told that their systems and plans have achieved a very high
level of visibility and excitement in their parent agency, the Dept. of
Interior, although the current size of the effort is not large.
Does this discussion answer the question of what a PC expert system can
do? Obviously not, but I hope it adds more light than heat to the current
discussions. In any case, a lot of the value of an expert system lies in the
knowledge base; I agree with Karnicky's observation that, regardless of
implementation, it's a long way from choosing red wine to discovering an
ore deposit.
Cheers,
Peter
------------------------------
Date: Tue 14 May 85 12:41:00-PDT
From: Carol Wright/Susie Barnes <WRIGHT@SUMEX-AIM.ARPA>
Subject: Seminar - Diagnosing Multiple Faults (SU)
SIGLUNCH
DATE: Friday, May 17, 1985
LOCATION: Braun Audiltorium, Mudd/Chemistry Building
TIME: 12:05
SPEAKER: Johan Dekleer
Xerox
TITLE: Diagnosing Multiple Faults
Diagnostic tasks require determining the differences between a model of
an artifact and the artifact itself. The differences between the
manifested behavior of the artifact and the predicted behavior of the
model guide the search for the difference between the artifact and its
model. The diagnostic procedure presented in this paper reasons from
first principles, inferring the behavior of the composite device from
knowledge of the structure and function of the individual components
comprising the device. The novel contributions of this research are:
Multiple-faults: No presupposition is made about the number of failed
components.
Measurements: Proposes optimal measurements to localize the fault.
Probabilistic: A priori probabilities of component faultedness are taken into
account.
Intermittency: The approach is robust in response intermittent faults.
The system is based on incorporating probabilistic information into an
Assumption-Based Truth Maintenance System.
------------------------------
Date: 10 May 1985 1429-EDT
From: Lydia Defilippo <DEFILIPPO@CMU-CS-C.ARPA>
Subject: Seminar - Fixpoint Extensions of First-Order Logic (CMU)
APPLIED LOGIC SEMINAR
Speaker: Yuri Gurevich (University of Michigan)
Date: Wednesday, May 15
Time: 2:00 - 3:15
Place: 5409 WeH
Topic: Fixpoint extensions of first-order logic
In 1979 Aho and Ullman noted that the relational calculus is unable to express
the transitive closure of a given relation, and suggested extending the
relational calculus by adding the least fixed point operator. From the point
of view of the expressive power, the relational calculus is exactly first-order
logic. Aho and Ullman's paper triggered an extensive study of the expressive
power of fixpoint extensions of first-order logic with emphasis on finite
structures. We survey the results of this study.
------------------------------
Date: 10 May 1985 1433-EDT
From: Lydia Defilippo <DEFILIPPO@CMU-CS-C.ARPA>
Subject: Seminar - A New "Turing" Thesis (CMU)
JOINT LOGIC COLLOQUIUM
Speaker: Yuri Gurevich (University of Michigan)
Date: Thursday, May 16
Time: 2:00 - 3:30
Place: 4605 WeH
Topic: A new thesis
Turing's thesis is that every computable function can be computed by an
appropriate Turing machine. The informal proof of the thesis gives more:
every computing device can be simulated by an appropriate Turing machine. The
following much stronger form of the thesis seems to be very much accepted
today: every sequential computing device can be simulated by an appropriate
Turing machine in polynomial time.
Turing machines are computational devices with unbounded resources. First, we
adapt Turing's thesis to the case when only devices with bounded resources are
considered.
Second, we define a more general kind of abstract computational devices, called
dynamic structures, and put forward the following new thesis:
Every computational device an be simulated by an appropriate dynamic structure
- of approximately the same size - in real time: a uniform family of
computational devices can be uniformly simulated by an appropriate family of
dynamic structures in real time.
In particular, every sequential computational device can be simulated by an
appropriate sequential dynamic structure.
A contribution of Andrea Blass is acknowledged. Descriptions of computational
devices are solicited for further confirmation of the thesis.
------------------------------
Date: 7 May 1985 1140-EDT
From: Lydia Defilippo <DEFILIPPO@CMU-CS-C.ARPA>
Subject: Course - Model Theory (CMU)
[Forwarded from the CMU bboard by Laws@SRI-AI.]
Department of Mathematics
CARNEGIE-MELLON UNIVERSITY
COURSE ANNOUNCEMENT - MODEL THEORY
Instructor: Ana Pasztor
Textbook: Model Theory, by Chang and Keisler, North Holland, 1973
When: Fall 1985, MWF - 2:30-3:20, Porter Hall 125C
Note: If you would like to take this course and have conflicts with the
time, PLEASE LET ME KNOW. (Room 7219 or phone x2558)
Course No., Credit, and Grade: 21-753, 12 Units, based on homework.
Aimed at: those who have had a basic course in logic and are interested in
broadening their knowledge.
What is model theory:
Model theory is the branch of mathematical logic which deals with
relation between a formal language and its interpretations, or models. We
shall study the model theory of first order predicate logic.
Models are structures of the kind which arise in mathematics or
computer science. To arrive at a model theory, we set up our formal language
of first-order logic by specifying a list of symbols and giving rules by which
sentences can be built up from the symbols. The reason for setting up a formal
language is that we wish to use the sentences to reason about the models.
Typical results of model theory say something about the power of
expression of first-order predicate logic. Lowenheim's theorem, for example,
shows that no consistent sentence can imply that a model is uncountable, or
Morley's theorem shows that first order predicate logic cannot, as far as
categoricity is concerned, tell the difference between one uncountable power
and another.
Model theory also gives methods of constructing models. A special
attention will be given in this course to ultraproduct constructions and their
applications in mathematics and computer science.
Much of model theory deals with the interplay of syntactical and
semantical ideas.
------------------------------
End of AIList Digest
********************