Copy Link
Add to Bookmark
Report
AIList Digest Volume 4 Issue 212
AIList Digest Friday, 10 Oct 1986 Volume 4 : Issue 212
Today's Topics:
Query - Integer Equations,
Expert Systems - Mathematical Models,
Philosophy - Man's Uniqueness & Scientific Method &
Understanding Horses & Irrelevance of Searle's Logic
----------------------------------------------------------------------
Date: Tue 7 Oct 86 10:36:01-PDT
From: Ken Laws <Laws@SRI-STRIPE.ARPA>
Subject: Integer Equations
RIT Researchers Find Way to Reduce Transmission Errors,
Communications of the ACM, Vol. 29, No. 7, July 1986, p. 702:
Donald Kreher and Stanislaw Radziszowski at Rochester Institute of Technology
have discovered a new geometry, the third 6-design, non-Euclidean geometry,
that allows solution of difficult problems in designing error-correcting
transmission codes. One problem with 99 integer equations and 132 unknowns
was solved in 12 hours; previous search methods would have required several
million centuries.
Integer (Diophantine) equations are notoriously difficult to solve. Is this
a breakthrough for other problem domains where search is used (e.g., bin
packing, traveling salesman, map coloring, and the "approximately-solved"
algorithms)? Is it a form of linear programming?
-- Ken Laws
------------------------------
Date: 6 Oct 1986 13:08:20 EDT
From: David Smith <DAVSMITH@A.ISI.EDU>
Subject: Expert systems and deep knowledge
Grethe Tangen asked about using mathematical models of gas turbines
as deep knowledge sources for diagnostics. GE in Schenectady, NY
are working in this area. Bruce Pomeroy is perhaps the best contact,
and he can be reached by mail to SWEET@a.isi.edu, or by phone
at (518)387-6781. Hope this helps.
DMS
------------------------------
Date: 7 Oct 86 15:52:00 GMT
From: mcvax!unido!ztivax!bandekar@seismo.css.gov
Subject: Mathematical Models
I see some difficulties in using mathematical models of technical systems
as a source of deep knowledge. Mathematical models are usually derived
from the structural information about the devices, and one particular mo-
del can represent more that one physical device. But I guess the approach
would not be impossible as long as you can derive your device structure
from your mathematical model. For example transfer function of several devices
may be mathematically expressed in the same way. For multiple input/output
plants the choice of state variables varies for state space representation.
Which variables are affected if a particular physical component is defective
and the causal ordering of the variables could be a valuable piece of know-
ledge for the purpose of diagnosis. Here, if you can map your model into
structural equations you may compute the causal ordering of the state
variables.[Iwasaki,Simon '86]. Hierarchical representation of the
technical systems is always useful. The concept of views[Struss, 86
to be presented at Sydney Univ. during Feb. 1987] is also important.
If you can tell me more about your problem, I may be able to help out.
my address: ... unido!ztivax!bandekar
Vijay Bandekar
:w
:q
------------------------------
Date: Mon, 6 Oct 86 10:41:45 EDT
From: "Col. G. L. Sicherman" <colonel%buffalo.csnet@CSNET-RELAY.ARPA>
Subject: man's godlike
I'm amazed that nobody has responded to Peter Pirron's last argument:
> The belief, that man's cognitive or intelligent abilities will
> never be reached by a machine, is founded in the conscious or
> unconscious assumption of man's godlike or godmade uniqueness,
> which is supported by the religious tradition of our culture. It
> needs a lot of self-reflection, courage and consciousness about
> one's own existential fears to overcome the need of being unique.
> I would claim, that the conviction mentioned above however
> philosophical or sophisticated it may be justified, is only the
> "RATIONALIZATION" (in the psychoanalytic meaning of the word) of
> understandable but irrational and normally unconscious existential
> fears and need of human being.
Even net.ai, which is still a chaos of wild theories, has gone beyond
regarding the a.i. question as a matter of science versus religion.
Some arguments against Pirron's conjecture:
-- If the objection to a.i. is rooted in cultural dogma, it's illogical
to look at the psychology of the individual. Every individual is, now
and always, unique--though some of us may feel that we are too much
like others. This is quite another question than whether our species
is unique.
-- Other animals, and even plants, have intelligence--not to mention
viruses! Many of us regard even a dog's intelligence as beyond the
capabilities of a.i., at least in the way that scientists presently
think about a.i.
-- Even an electric-eye door can be regarded as a successful implementation
of artificial intelligence. We skeptics' greatest doubts tend to focus
on theories of emergent intelligence--theories as attractive to some
modern researchers as the Philosopher's Stone was to medieval researchers,
and (some say) with just as little basis in the nature of things.
-- To divide intelligent beings into men and machines is not necessarily
precise or exhaustive. For example, ghosts may be intelligent
without belonging to either category.
-- A secular equivalent of "godlike uniqueness" is that man is special:
that we mean more to ourselves than does anything else, living or lifeless.
Only a scientist would argue with this. 8 |-I
------------------------------
Date: 9 Oct 86 05:04:59 GMT
From: allegra!princeton!mind!harnad@ucbvax.Berkeley.EDU (Stevan Harnad)
Subject: Re: Turing test - the robot version
>>> instead of a computer trying to fool you in ASCII,
>>> it's a robot trying to fool you in the flesh...
>>> Remember, scientists aren't just trying to make things better for you.
>>> They're also trying to fool you!
The purpose of scientific inquiry is not just to better the human
condition. It is also to understand nature, including human nature.
Nothing can do this more directly than trying to model the mind. But
how can you tell whether your model is veridical? One way is to test
whether its performance is identical with human performance. That's no
guarantee that it's veridical, but there's no guarantee with our
models of physical nature either. These too are underdetermined by
data, as I argue in the papers in question. And besides, the robot
version of the turing test is already the one we use every day, in our
informal solutions to the other-minds problem.
Finally, there's a world of difference, as likewise argued in the
papers, between being able to "fool" someone in symbols and being able
to do it in the flesh-and-blood world of objects and causality. And
before we wax too sceptical about such successes, let's first try to
achieve them.
Stevan Harnad
princeton!mind!harnad
------------------------------
Date: 10 Oct 1986 06:39 EDT (Fri)
From: Wayne McGuire <Wayne%OZ.AI.MIT.EDU@XX.LCS.MIT.EDU>
Subject: Understanding Horses
Date: Mon 29 Sep 86 09:55:11-PDT
From: Pat Hayes <PHayes@SRI-KL.ARPA>
Subject: Searle's logic
Look, I also don't think there's any real difference between a human's
knowledge of a horse and [a] machine's manipulation of the symbol it is
using to represent it.
At one end of the human knowledge spectrum we have that knowledge of a
horse which is aware that two horses + two horses = four horses; at the
other end is that sort of rich and unfathomably complex knowledge which is
expressed in a play like Peter Shaffer's _Equus_, and which fuses, under
the force of sympathetic imagination, conceptual, emotional, biological,
and sensorimotor modes of cognition. I suppose that our most advanced
expert systems at the elementary end of the cognitive spectrum can capture
knowledge about the structural and functional features of a horse, but it
is not clear that any knowledge representation scheme will EVER simulate
what is most interesting about human cognition and which relies on
unconscious and intuitive resources. In one dimension of cognition the
world is a machine, an engineering diagram, which is readily accessible by
bit twiddling models; in another, that of, say, Shakespeare, it is a living
organism, whose parts are infinitely interconnected and partially decrypted
only by the power of the imagination. And so I would argue, with regard to
human and machine cognition of horses or anything else, that there is a
major difference in any dimension of knowledge that counts, and that
repairing automobiles or space stations, and writing or understanding poems
(or understanding the world in the broadest sense), have nearly nothing in
common.
Wayne McGuire
(wayne@oz.ai.mit.edu)
------------------------------
Date: Fri, 10 Oct 86 11:57:31 edt
From: Mike Tanner <tanner@ohio-state.ARPA>
Reply-to: tanner@osu-eddie.UUCP (Mike Tanner)
Subject: Re: Searle's logic
Pat Hayes made some cogent remarks about Searle's problems with AI
being much deeper than the discussion here would indicate. But I
wonder whether the argument is worth the effort.
I have a lot of work to do and only so much time. I can work just
fine on problems of intelligence without worrying about Searle's (or
Dreyfus's) complaints. Just as the working physicist can work all day
without once being bothered by the question of whether quarks *really*
exist, so the working AIer can make progress on his problems without
being bothered by Searle.
-- mike
tanner@ohio-state.arpa
------------------------------
End of AIList Digest
********************
