Copy Link
Add to Bookmark
Report
AIList Digest Volume 5 Issue 099
AIList Digest Tuesday, 14 Apr 1987 Volume 5 : Issue 99
Today's Topics:
Queries - Statistical Expert Systems & Prolog for UNIX System V.2 &
Cog Sci Conference & Cog. Psych. Grad. Schools,
AI Tools - MS-DOS Expert System Tools,
Application - AI in Network Protocols,
Funding - Military Funding,
Humor - Demon Error & Which Half is Right?,
Inference - Clyde the Elephant
----------------------------------------------------------------------
Date: Mon 13 Apr 87 16:19:12-PST
From: Ken Laws <LAWS@IU.AI.SRI.COM>
Subject: Statistical Expert Systems
I have received a letter from Dr. D.J. Hand, Institute of Psychiatry,
De Crespigny Park, Denmark Hill, London, SE5 8AF. He (or she?) is
trying to compile a list of researchers working in statistical expert
systems, for use by researchers and conference organizers. If you
would like to be listed, and to receive a copy, send your name,
address, and a brief description of your work.
-- Ken
------------------------------
Date: 13 Apr 87 16:15:39 GMT
From: husc8!edwards@husc6.harvard.edu (Bill Edwards)
Subject: Good Prolog Interpreter/Compiler for UNIX System V.2
Wanted: Good Prolog Interpreter/Compiler for UNIX System V.2
Please respond by email-thanks. -- Bill Edwards
Bill Edwards edwards@harvard.harvard.edu (ARPA)
UNIX Systems Programmer/Analyst ...!harvard!edwards (UUCP)
Harvard Science Center edwards@harvunxu (BITNET)
1 Oxford Street hucsc::edwards (DECNET)
Cambridge, MA 02138
------------------------------
Date: Mon 13 Apr 87 16:53:40-EDT
From: John C. Akbari <AKBARI@cs.columbia.edu>
Subject: cog sci conference
anyone have an email address to inquire about attending the cognitive science
conference in july (just after aaai-87).
john c akbari
ARPANET & Internet akbari@CS.COLUMBIA.EDU
BITnet akbari%CS.COLUMBIA.EDU@WISCVM.WISC.EDU
uucp & usenet ...!seismo!columbia!cs!akbari
DECnet akbari@cs
PaperNet 380 riverside drive, no. 7d
new york, new york 10025 usa
SoundNet 212.662.2476
------------------------------
Date: 10 Apr 87 17:19:04 GMT
From: seger@husc4.harvard.edu (carol seger)
Subject: Cog. Psych. Grad. School advice sought.
I am planning to apply to graduate schools in cognitive psychology /
cognitive science in the fall. However, I will be teaching
high school in Kenya for a year beginning in July, so I
have to decide where I want to apply soon so I can visit places before
I leave. I am seeking three sorts of advice:
a. Other schools that might be worthwhile but that I have overlooked
b. Any information from current students in any of the programs I
am considering.
c. General advice on the applications process. As far as I am aware,
none of the programs I am applying to require interviews. Is
there any reason I cannot apply from overseas?
I am interested in high-level perception, natural reasoning,
categorization, spatial cognition, cognitive development and cognitive
neuropsychology. I am not interested in psycholinguistics (at least at
the moment) or low-level perception. I prefer to concentrate on
experimental psychology -- while I find computer modeling to be
interesting, I don't want to do it myself.
I am currently a senior in the cognitive science option of the
Psychology concentration at Harvard Univeristy. I wrote my senior
thesis on intermodality realtions in shape perception, and have worked
on a naturalistc study of mental imagery.
So far, in order of preference, I am applying to
Stanford
UC Berkeley
UC San Diego
UCLA / University of Pennsylvania (tie, so far).
I prefer to live in California, but, of course, I'll go to the best
program I can get into.
If you have any advice, please mail it to me. Many thanks.
Carol Seger.
carol@borax.lcs.mit.edu
seger@wjh12.harvard.edu
------------------------------
Date: 5 Apr 87 04:46:06 GMT
From: nbires!isis!csm9a!japplega@ucbvax.Berkeley.EDU (Joe Applegate)
Subject: Re: MS-DOS expert system tools?
>
>I'm looking for expert-systems tools that can be run on PC-class machines
>
>So far I have looked at GURU, EXSYS, VP-Expert, and KDS. None of these
>systems comes close to my needs; most of them are question/answer menu-based
>tools best suited for very simple interactive diagnostic or recommendation
>ES's. Some of them allow access to external databases, but none of them
>(as far as I can tell) allow general user routines to be linked in.
>
>I have access to TI's PC-Plus and will be looking at it soon. I have read the
>advertising blurb on Level Five's Insight-2+, and it sounds very interesting.
>But then, the blurbs on some of the other tools sounded good, too.
>
>Has anyone used these systems, or any others that meet my needs? I would
>really appreciate it if you would contact me with any suggestions.
>
Most PC based Expert Systems shells are oriented towards database type
type queries of their knowledge. Though it is possible to acces both ports
and the bios from TI's Personal Consultant Plus, I doubt if that or any shell
will give the response needed for real time processing.
A more feasible method for this type of development is to do it from scratch
in an acceptable language... most probably Lisp or Prolog, though C and
Pascal can be used in such an environment and have been in the past!
At the risk of getting lynched I would recommend you take a look at Turbo
Prolog... if your rule base is not dynamic, Turbo Prolog provides a powerful
yet inexpensive development engine with graphic primitives and direct access
to DOS and BIOS functions as well as the I/O ports of a PC.
Joe Applegate - Colorado School of Mines Computing Center
{seismo, hplabs}!hao!isis!csm9a!japplega
or
SYSOP @ (303) 273-3989 300/1200/2400 8-N-1
Minds of Mines AI BBS
------------------------------
Date: 7 Apr 87 13:21:24 GMT
From: sundc!cos!duc@seismo.CSS.GOV (Duc Kim Nguyen)
Subject: Re: AI in Network Protocols.
I think this is a very interesting topic for discussion.
Typically a protocol specification contains some BNF notation
for the syntactical definition (e.g., X.400, etc...) and the binding
(or usage/meaning) of the components' values is burried in the 'english'
text of the spec. The effect of this is the lack of a more 'complete'
and/or formal notation to capture both the syntax and semantic in order
for automating a testing system (to test the protocol) and/or
determining a set of test cases to be 'partially' (or even wholly)
complete to test a set of functionalities of the protocol (and
therefore a result analysis system can be automated).
Maybe, a knowledge-based system will solve this, but I
prefer not to think about a database-driven approach until no can-do.
Duc Kim Nguyen
Corporation for Open Systems
------------------------------
Date: 5 Apr 87 06:28:07 GMT
From: ubc-vision!calgary!vuwcomp!steve@seismo.CSS.GOV (Steve Cassidy)
Reply-to: steve@vuwcomp.UUCP (Steve Cassidy)
Subject: Military Funding
In article <[A.ISI.EDU]31-Mar-87.15:25:11.DAVSMITH> DAVSMITH@A.ISI.EDU writes:
>Without the military applications, who in the commercial sector
>would attempt to put together cooperating expert systems
>in real-time? [ One could broaden the issue and ask
>"Who in their right mind would..?"]
Here we assume that the only *possible* applications of real-time cooperating
ES are military ones. What is the major difference between a system which sits
in a fighter plane monitoring the pilots actions and one which sits in some
complex manufacturing plant monitoring the processes there?
Too often military research is justified as the only way new ideas can
develop, the truth is that they are the only research programmes given
sufficient funds to develop new ideas. If research groups had the same level of
funding available for civil projects then they would be able to develop
real-time cooperative expert systems in domains which may actually be *useful*
to mankind.
Steve
ACSnet: steve@vuwcomp.nz
UUCP: {ubc-vision,alberta}!calgary!vuwcomp!steve
------------------------------
Date: Mon, 13 Apr 87 01:33:15 pst
From: Eugene Miya N. <eugene@ames-pioneer.arpa>
Subject: Feigenbaum Comment about SDI
Last week, Computer Literacy bookstore completed a kick off opening
lecture series, some very noted computers scientists spoke over the
course of two weeks. I think comments by two speakers would interest
recent discussions on both Arms-d and AIlist. In particular,
Dr. Ed. Feigenbaum made mention in his words that the SDIO has
dropped funding of AI from its budget. His implication appeared
to be more software engineering oriented rather than battle
management oriented. I don't know if they have or not, but I personally
do not get the impression that they have, especially for battle
management.
>From the Rock of Ages Home for Retired Hackers:
--eugene miya
NASA Ames Research Center
eugene@ames-aurora.ARPA
"You trust the `reply' command with all those different mailers out there?"
"Send mail, avoid follow-ups. If enough, I'll summarize."
{hplabs,hao,ihnp4,decwrl,allegra,tektronix,menlo70}!ames!aurora!eugene
------------------------------
Date: 6 Apr 87 17:48:27 GMT
From: "Col. G. L. Sicherman"
<colonel%sunybcs%math.waterloo.edu@RELAY.CS.NET>
Subject: Re: AIList Digest V5 #92
In article <MINSKY.12290784989.BABYL@MIT-OZ>, MINSKY@OZ.AI.MIT.EDU writes:
> The term "demon" comes from Oliver Selfridge, via the paper,
> "Pandemonium: A Paradigm for Learning", published in Symposium of the
> mechanization of Thought Processes, November 1858.
Then we can certainly concede priority to Selfridge. I wonder how much
influence he had on the work of Babbage? (Ken, you should have caught this!)
--
Col. G. L. Sicherman
UU: ...{rocksvax|decvax}!sunybcs!colonel
CS: colonel@buffalo-cs
BI: colonel@sunybcs, csdsiche@ubvms
------------------------------
Date: 13-Apr-1987 0938
From: kevin%bizet.DEC@decwrl.DEC.COM (Now, if it sounds good, you
don't worry what it is: you just go and enjoy it.)
Subject: Which half is right?
> [I have faith that only half of what I know is wrong. I'll
> let you know when I find out which half. -- KIL]
Merely determining the halves, let alone figuring out which half is right, would
be an astonishing accomplishment!
Kevin LaRue
------------------------------
Date: 5 Apr 87 20:36:23 GMT
From: "Col. G. L. Sicherman"
<colonel%sunybcs%math.waterloo.edu@RELAY.CS.NET>
Subject: Re: Clyde the elephant
The problem of Clyde the elephant brings up one of the biggest
controversies in statistics, one which is starting to spill over
into A.I. To recapitulate:
1. 95% of elephants are grey;
2. 40% of royal elephants are yellow;
3. Clyde is a royal elephant.
But we know nothing about what percentage of elephants are royal.
The distribution could look like this:
| royal common
------+---------------
grey | 15 175
yellow| 10 0
or like this:
| royal common
------+---------------
grey | 0 95000
yellow| 2 0
red | 3 4995
Can we assign a valid probability to "Clyde is grey" without knowing
the likelihood of either distribution (or any other)? One school of
thought says no--the best we can do is follow Boole's suggestion of
computing upper and lower bounds for the probability. Other schools,
notably that led by A. P. Dempster, say yes.
And this topic is all too philosophical enough to be discussed here
in mod.ai!
--
Col. G. L. Sicherman
UU: ...{rocksvax|decvax}!sunybcs!colonel
CS: colonel@buffalo-cs
BI: colonel@sunybcs, csdsiche@ubvms
[SRI is a hotbed of Dempster-Shaferism, so I'll take a chance on
clarifying this. Tom Garvey or other readers can correct me if
I'm off base. The Dempster-Shafer (D-S) is to track upper and
lower bounds for probability. This is controversial in two ways:
Dempster's rule for combining contradictory evidence, and the
power/appropriateness/usefulness of the interval approach in general.
(Conflicting evidence really doesn't enter into the Clyde problem.)
It is the Bayesians who generally assign probabilities, although
they don't do it is blindly as their "loyal opposition" would
imply -- while underlying uniform or even Gaussian distributions
are typically assumed for predictive power under random sampling,
Bayesians might choose a "pessimal" a priori distribution to model
tricky situations such as this one. They can also do symbolic Bayesian
analysis with free parameters in order to derive formulas that are
valid for any state of the world. Fuzzy logicians use a very similar
theory, but are likely to assume that the underlying distributions
are typically implied by the manner in which the problem is stated.
A fourth group, perhaps led by Tversky and Kahneman, are more interested
in the analogy-based reasoning of humans than in optimal decision
theory. And others, e.g. Cohen and various expert systems researchers,
are willing to consider any type of estimate as long as the justification
is given (for use in further reasoning).
Intervals are nice because they make no unwarranted statements.
(Disclaimer: The endpoints may themselves by subject to sampling
errors. Logic-based methods, including D-S, can be very sensitive
to errors in the intial evidence -- as can methods based on tightly
constrained a priori distributions.) Upper and lower probabilities
are also more informative than single point estimates, and can
be interpreted as recording what is unknown as well as what is
known. In cases where a parametric distribution is appropriate,
however, the parameters of that distribution (or optimal estimates
thereof) are the most powerful estimates of the state of the world.
Intervals are not convenient for representing true Gaussian distributions,
for instance, since the intervals must be infinite in extent. (One
might want to use intervals for the mean and standard deviation, though.)
I tend to believe that all sampled data is Gaussian unless there is
evidence to the contrary (either a priori or from examination of the
data), partly because that leads to points estimates and distributions
thereof that are useful. I would not attempt to impose this assumption
on Clyde, however, and there are many situations calling for non-Bayesian
reasoning. -- KIL]
------------------------------
End of AIList Digest
********************
