Copy Link
Add to Bookmark
Report
AIList Digest Volume 2 Issue 096
AIList Digest Friday, 27 Jul 1984 Volume 2 : Issue 96
Today's Topics:
AI Culture - Genealogy of AI,
AI Literature - Robotics Directory,
AI Languages - Lisp Speed Benchmarks,
Humor - Naming Names,
Review - Neuroanatomy and Electromagnetic Waves,
Turing Test - Discussion
----------------------------------------------------------------------
Date: 23 Jul 84 15:41:57-PDT (Mon)
From: hplabs!hpda!fortune!amd!decwrl!flairvax!pfps @ Ucb-Vax.arpa
Subject: Genealogy of AI
Article-I.D.: flairvax.677
David S. Johnson of ATT Labs has gathered together a genealogy of theoretical
computer science along intellectual, rather than biological bloodlines. That
is, the parent-child relationship has been replaced by the more significant
PhD advisor-advisee relationship. I have just seen a report that he produced
that contains 672 of these entries and several genealogical trees taken from
the data and I thought that it would be a nice idea to produce a similar
listing for Artificial Intelligence. This data could show how AI has spread
from its initial centres to its current broad coverage.
One problem with such a genealogy is collecting and organizing all the data.
Therefore I am asking for anyone who wants to contribute data about AI
advisor-advisee relationships to send mail to me. To make the organizing
process easier I would like all responders to follow the strict format
detailed below:
1/ no ``bug killer'' lines
2/ each entry is one line and should contain the following information
a) advisee name with surname first
b) advisor name with surname first
c) institution where degree granted (in a short format)
d) year in which degree granted (all four digits)
e) type of degree (PhD, MSc, or other graduate degree)
f) field of research (AI, Physics, Mathematics, etc.)
f) area of research (natural language, expert systems, etc.)
g) current affiliation of advisee (in a short format)
3/ fields separated by # characters
4/ unknown values indicated by ? characters
5/ null values indicated by empty fields
6/ all entries together at the beginning of the message and followed
by a blank line
Here are three sample entries:
Cohen, Robin#Perrault, Ray#Toronto#1984#PhD#AI#natural language#Waterloo
Patel-Schneider, Peter F.#Mylopoulos, John#Toronto#1978#MSc#AI#knowledge
representation#FLAIR
Patel-Schneider, Peter F.#McCalla, Gord#Toronto#1978#MSc#AI#knowledge
representation#FLAIR
[I had to break the longer lines for AIList. -- KIL]
When entering names try to use the name from the thesis unless the person's
name has since changed and the changed version is more well-known. For
institution names try to use the shortest name in common use which is unique.
For example, use Toronto for University of Toronto, FLAIR for Fairchild
Laboratory for Artificial Intelligence, and other well-known short forms such
as Berkeley and MIT but NOT UofT for the University of Toronto or the
University of Texas. Fields of research include AI, Computer Science,
Physics, Mathematics, Philosophy, Psychology, and Chemistry. The idea here is
to find out the backgrounds of AI people which is why AI is separated from
Computer Science. Areas of research are most important when the field is AI
and areas within AI include but are not limited to natural language, knowledge
representation, expert systems, theorem proving, learning, and vision.
A thesis advisor is that person officially recognized by the granting
University as the advisor. If there is no such person then the advisor is
that person who guided the research for the thesis. Such an advisor should be
indicated by a '*' character after the name. If the official advisor did
nothing besides act as a signing agent and someone else really did all the
work then include two entries for the thesis, one listing the official advisor
and the other the unofficial advisor again with a trailing '*' character.
Also include two entries if there were two official advisors or two unofficial
advisors but please do not go beyond two advisors for one thesis.
To keep the amount of data within reasonable limits I am really only
interested in people who are in AI (preferably doing research) or who have
advised (directly or indirectly) someone in AI. So if you are in AI the data
concerning you that I am interested in are your thesis advisor(s), their
advisors, and so on as far back as can be traced. Of course, you can also
include other relevant data if you so wish. If you know for certain that some
advisor has no advanced degree please include this. I will assume that if
someone has only a master's degree listed then that person has no PhD.
I will collect all information sent to me and do as much error correction and
redundancy elimination as possible. If enough responses are generated I will
send out periodic lists of the information generated, otherwise I will reply
only to the respondants.
Peter F. Patel-Schneider {decvax!decwrl,hplabs}!flairvax!pfps
------------------------------
Date: 24 Jul 84 12:50:03-PDT (Tue)
From: ihnp4!houxm!mhuxl!ulysses!unc!mcnc!ncsu!uvacs!edison!rmk @ Ucb-Vax.arpa
Subject: Robotics Directory
Article-I.D.: edison.317
A few weeks ago someone posted a notice about the 1984 edition of the
International Robotics Industry Directory. Would someone tell me
the publisher of this directory, or where it could be obtained?
Thanks much in advance,
Bob Kossey
{...houxm,...decvax!mcnc!ncsu!uvacs!iedl02}!edison!rmk (804) 978-6378
GE - Industrial Electronics Development Lab Charlottesville, Va.
------------------------------
Date: 25 Jul 1984 12:31:23-EDT
From: Philip.Kasprzyk at CMU-RI-ISL2
Subject: Lisp Speed Benchmarks
[Forwarded from the CMU bboard by Laws@SRI-AI.]
I am interested in determining the absolute speed of execution of a lisp
system. I realize that this is a complicated task that is application
and implementation dependant. Nevertheless I am seeking the following:
1) A "gibson mix" style benchmark for lisp based systems.
2) Any data that experienced users of lisp may have regarding performance.
(I don't care what machine or dialect of lisp is reported)
3) Any method or program which translates a lisp program into another
target high level language with execution speed as its objective.
(I don't mean a normal lisp compiler)
Can anyone out there help me?
If so send mail to pmk@isl2.
------------------------------
Date: Mon 23 Jul 84 09:54:56-CDT
From: David Throop <LRC.Throop@UTEXAS-20.ARPA>
Subject: naming names
I have started naming names.
It started with the observation that since I am David Throop, and David
Throop is my name, then it follows that I am my name. There is obviously a
flaw here; I am not my name. I mean, if I changed my name I would still be
myself. So it is perhaps more clear to say that I am David Throop, and my
name is "David Throop". (This is still off. I mean, "David Throop" is still
a character string, a sequence of letters, and my name is something other than
a sequence of letters. "David Throop" has a value that is a name, and the
name has a value that is a person, and I am that person.) [Are you still with
me?]
That is, to say of course, that D - a - v - i - d - - T - h - r - o - o - p
is still another character string that denotes a character string that denotes
a name that denotes me. (The string "me" denoting the same person,
coincidently, as is denoted by my name.)
Which is all prologue to the question of whether I could give my name a
name. Not the name "David Throop", of course. That's already taken. I
considered naming my name "david throop" but I felt that this might cause some
confusion. (And raises the ugly question of how to pronounce it. You see,
the "h" silent in my last name, and though "Throop" is pronounced with a
silent "h" I'm not sure that "throop" would be also.)
[Which brings up the side issue of the version of my name as a pronounced
set of sounds. And on reflection, I'm not sure whether the value of the
character string "David Throop" is a sound sequence, or my name itself. Or it
may be that the value of the sound sequence is the character string. Or its
more likely that the sound sequence and the character string are two separate
objects that happen to have the same value. Although, curiously, you can get
from one to the other and back again without ever encountering me. I mean,
even if you didn't know me, even if you didn't know that "David Throop" was a
name, you could pronounce it and if you heard it you could spell it. But
you'd probably have a little trouble with the silent "h". It persists in
injecting itself into the whole problem.]
But back to giving my name a name. People always say, "Well I'm not about
to start naming names", and I think the forgoing illustrates some of the
problems away from which people are shying. But then, how are we going to
talk about my name if it is nameless?
For instance, if I tell you that I don't want to sully my good name, and you
reply, "What good name?" how can I reply? If I reply, "Why, David Throop, of
course," then I haven't refered to my name, I've refered to myself. Of
course, I could reply "Why, "David Throop", of course" but those little quote
marks are kind of hard to see in a spoken retort, and that's the kind of
challenge I reply to immediately. It wouldn't do to have a letter show up a
week later saying "Why, "David Throop", of course." One needs to defend one's
name promptly. Some people have a cute way of waving their hands in the air
in order to indicate those marks, but it kind of takes the force out of the
retort, and I must remember that my good name is on the line.
So I've decided to name my name something else. Although I saw some good
ideas in a book named "Your Baby's Name", I steered clear. "Jason" sounded
nice, but somebody might think that that was somebody's name. And it's not.
It's a name's name. I decided on G00483; as near as I can tell it's not being
used for anything else right now. And it sounds like a name's name.
But this brings up a question. Is G00483 my name's name? Or is it just my
name for my name? (my own name, that is.) After all, my name doesn't have any
need for its name. I'm the one that needs to know its name, so that I can
refer to it when you question my good name. Since it doesn't name itself and
I do, I'll just leave it as my name for the name of myself.
Look, I realize this is all rather complex and I don't want to run it into
the ground. Just understand. I've started naming names. I've got a good
one. And for now, I'm retaining custody.
Sincerely,
David Throop
------------------------------
Date: Sun 22 Jul 84 23:34:17-PDT
From: Ken Laws <Laws@SRI-AI.ARPA>
Subject: Neuroanatomy and Electromagnetic Waves
The August issue of High Technology has an interesting article about
EEG potentials in the brain in response to different stimuli. It points
out, for instance, that the spatial patterns of the potentials are
very different for a subject hearing the noun "rose" and another hearing
the verb "rows". These patterns of oscillatory activity may be simply
a side effect of neural processing or a fundamental information transfer
mechanism. (It is suggested that transfer of signals via electromagnetic
radiation may be faster than broadcasting via neural interconnections,
but I find that hard to believe. It is also suggested that the resonant
coupling of neural circuits may be a robust transmission mechanism in
an organ that is continually rearranging synapses and even losing neurons
[at about 50,000 per day, I think].) The temporal frequency spectra of
these patterns are also presented as fuzzy hash functions possibly
responsible for associative memory.
A box accompanying the article discusses the Boltzmann Machine, an
architecture based on neural models. Scott Fahlman (of NETL fame) and
Geoffrey Hinton are quoted. (Scott mentioned this work in an AIList
issue last year.) The Boltzmann machine apparently has stochastic
behavior even for deterministic inputs; this simplifies stochastic
analyses of its behavior.
-- Ken Laws
------------------------------
Date: 22 Jul 84 14:12:05-PDT (Sun)
From: hplabs!ames!eugene @ Ucb-Vax.arpa
Subject: Should The Turing test be modified with the times?
Article-I.D.: ames.427
I am not an AI expert, but I do know image processing and certain other
computationally intensive tasks which are 'easy' for humans. I also know
the original definition of the Turing test. I take issue
with statements that LISP machines are 'smarter' or 'better' than humans
for "subgoal" tasks. What I am wondering is "should the Test be modified
to Our times?"
I recall that Turing specified that a communication link such as a tty or
phone could be used [1930s]. Should this be changed to a video link?
[This is am example only, there might be other aspects.]
Should the testee 'see' images? Can machines recognize defocused images
of an animal and say "That is a cow" as humans could [to a limit].
Perhaps our definition of 'human' constitutes a moving target which
might make the Test more difficult. The processing requirements of the
Turing Test in the 1930s would be less than those of today. I can see it now:
over a crude link, we discover that we cannot tell the difference between
man and machine, then we hook up a video link and the difference 'becomes
apparent.' Admittedly, one can argue that this is only a matter of adding
more processig power, but ignore that argument for a while. Also, there might
be audio examples (perhaps not as powerful as the video example).
Comments? This is for discussion purposes, not just me.
--eugene miya
NASA Ames Research Center
------------------------------
Date: 22 Jul 84 10:15:25-PDT (Sun)
From: decvax!mit-athena!yba @ Ucb-Vax.arpa
Subject: Re: Re: The Turing Test - machines vs. p - (nf)
Article-I.D.: mit-athe.213
<Dave Seaman, re: phd thesis and orals passing>
You are trivializing the point, not understanding it. The test of whether
an engineer understands engineering is that he can design and build things
that work. I have met Ph.D.s who can write "learned" papers but cannot
"do" anything concrete.
If you must have a test, I'll assert that someone who can apply knowledge
of a field to a new area, and can transmit that knowledge to another who
was previously ignorant of said knowledge such that that person can apply
the knowledge in the same way, has understanding (yes, it's recursive).
Measurement of AI performance is important. It is notions like "test"
that cause people to confuse "production systems" with "expert systems".
You may recall that the original notion of "expert system" was "a system
that solves problems the way (human) experts do". This has been reduced
to rules-based production systems in many people's minds, because they
think that experts solve problems by applying rules. I am not satisfied
that this is true; the question several people have asked is "does a
rules based system demonstrate an expert's intuition?" After all, it passes
the "test" of applying knowledge to a problem. You can substitute
"judgement", "intelligence", "understanding", or "talent" for the word
intuition if you like.
The question becomes rather concrete when you decide whether to allow
a program to practice medicine. We have all heard of examples of
accredited (human) doctors who have not been able to safely practice
medicine although they passed all the qualifying "tests".
In fact, there seems to be a great body of technique floating around in
many disciplines; there is also a great lack of those who know what the
limits of application of those techniques are (usually because they
know the underlying assumptions and constraints). I greatly fear
people who have become so proficient at using hammers that every problem
begins to resemble a nail.
I will also assert that you read my previous letter, processed the information,
responded, and did all this without understanding what I meant. Now if we
assume you disagree, what test would you design to see which of us was
correct? (Warning: this is a hard problem).
--
yba%mit-heracles@mit-mc.ARPA UUCP: decvax!mit-athena!yba
------------------------------
Date: 24 Jul 84 9:46:04-PDT (Tue)
From: pur-ee!CS-Mordred!Pucc-H.Pucc-I.ags @ Ucb-Vax.arpa
Subject: Re: Re: The Turing Test - machines vs. p - (nf)
Article-I.D.: pucc-i.374
Reply to Mark Levine:
The original example was of a program which passes calculus tests. You
objected that such a program has not really demonstrated understanding.
I agreed with your point before, and though I did not explicitly say so,
I thought it was implicit in the fact that I did not express disagreement.
I then suggested a different test: suppose a program writes a Ph.D.
dissertation and passes its "orals." I didn't mention any specific field,
but I was thinking of mathematics, where a Ph.D. dissertation involves the
development of new mathematics. I then asked whether this program has
demonstrated understanding of its field.
> If you must have a test, I'll assert that someone who can apply knowledge
> of a field to a new area, and can transmit that knowledge to another who
> was previously ignorant of said knowledge such that that person can apply
> the knowledge in the same way, has understanding (yes, it's recursive).
I submit that the test I suggested meets the first half of your criterion.
You have added a new point here which I overlooked: the ability to transmit
knowledge should also be considered part of the test. I don't believe this
part should be weighted as heavily, however, since the best doers are not
necessarily the best teachers. My objective was not so much to establish
the definitive test but to explore the question of whether a computer can
demonstrate understanding of a particular field (which is closely related
to the question of whether an adequate test can be devised). I don't
understand why you think this is "trivializing the point." You admitted
yourself that "measurement of AI performance is important."
>I will also assert that you read my previous letter, processed the information,
>responded, and did all this without understanding what I meant. Now if we
>assume you disagree, what test would you design to see which of us was
>correct? (Warning: this is a hard problem).
I really don't understand why you are being so defensive. I agreed with your
original point and I have already said so. In my own previous posting I did
not state any opinions; I merely posed a question. My objective was
enlightenment. I am sorry you interpreted this as an attack on your position.
--
Dave Seaman My hovercraft is no longer full of
..!pur-ee!pucc-i:ags eels (thanks to my confused cat).
------------------------------
End of AIList Digest
********************
