Copy Link
Add to Bookmark
Report
AIList Digest Volume 6 Issue 076
AIList Digest Thursday, 21 Apr 1988 Volume 6 : Issue 76
Today's Topics:
Opinion - Hot Research Topics & Goals of AI & Free Will
----------------------------------------------------------------------
Date: Thu, 14 Apr 88 11:11:35 HAE
From: Spencer Star <STAR%LAVALVM1.BITNET@CORNELLC.CCS.CORNELL.EDU>
Subject: Re: AIList V6 #69 - Queries
Exiting work in AI. There's lots of it. The three criteria
are: 1. Highly thought of by at least 50% in the field.
2. Positive contribution
3. Real AI
Machine learning is a real AI field; there is general agreement
that learning is central to real AI. Machine learning is perhaps
the major subfield devoted to AI learning, although most other
subfields also touch upon learning. Some, like neural networks,
are centered on learning.
Surprisingly, I don't see that much controversy in machine learning. There
is solid progress being made on several fronts. Recent controversies have
been on esoteric issues like whether there is a tradeoff between generalization
and efficiency, whether facts in the deductive closure of a system can be
said to be learned, etc. No big battles with rival camps raging at each
other.
There is, however, solid research. Hal Valiant and David Haussler have
made good theoretical progress at defining a certain type of learning.
Explanation-based learning is a very exiting hot!!! area for research
right now. At the Stanford Symposium many people made progress reports
on hybrid systems that use the deductive inference engine based on
PROLOG-EBG or EGGS or some variant, and then include inductive techniques
to do learning on both deductive and inductive levels. Another area
involves classification trees of the sort generated by Quinlan's ID3
program. There is wide agreement that this is a positive contribution.
And it is not a controversial technique.
SOAR is an architecture being worked on by people at severl universities.
Although the claims of the group have been controversial, the actual
work they are doing is well thought of. And copies of the program
are available to researchers for their own experimentation.
Take your pick. There is lots to choose from.
Spencer Star
------------------------------
Date: 18 Apr 88 14:08:45 GMT
From: hubcap!ncrcae!gollum!rolandi@gatech.edu (rolandi)
Subject: consensus in ai and psychology
In response to Ehud Reiter's:
>I was recently asked (by a psychology graduate student) if there was
>any work being done in AI which was widely thought to be exciting and
>pointing the way to further progress. Specifically, I was asked for work
>which:
> 1) Was highly thought of by at least 50% of the researchers in
>the field.
> 2) Was a positive contribution, not an analysis showing problems
>in previous work.
> 3) Was in AI as narrowly defined (i.e. not in robotics or vision)
>I must admit that I was (somewhat embarassingly) unable to think of
>any such work. All the things I could think of which have people excited
>(ranging from non-monotonic logic to connectionism) seemed controversial
>enough so that they could not be said to have the support of half of all
>active AI researchers.
Psychology itself would look pretty bad if asked the same sort of questions.
No discipline is more factionalized than psychology. Its representatives
range from scientific materialists to existential philosophers I don't
think you could get 50% of psychologists to even agree as to the proper
subject matter of their discipline.
Walter Rolandi
rolandi@gollum.UUCP
NCR Advanced Systems Development, Columbia, SC
University of South Carolina Departments of Psychology and Linguistics
------------------------------
Date: 18 Apr 88 14:19:46 GMT
From: mit-amt!mob%mit-amt.MEDIA.MIT.EDU@EDDIE.MIT.EDU (Mario O
Bourgoin)
Reply-to: mit-amt!mob%media-lab.MEDIA.MIT.EDU@EDDIE.MIT.EDU (Mario O
Bourgoin)
Subject: Re: Prof. McCarthy's retort
A better, cute and compact argument for Lisp: Scheme.
------------------------------
Date: Mon Apr 18 10:48:41 EDT 1988
From: sas@BBN.COM
Subject: Why AI? ---- Slightly humorous ----
I wonder if we can get the same arguments going about "business
programming". For example,
-----
- How can you tell if a program is a business program?
They deal with dollars (or other monetary units).
Hmm, so an employee assignment program which only deals with
people's names, times and office numbers is not a business
program.
How about, it deals with time or money?
So if we write a program to see who is elible for class IIB
promotions, but we don't actually check existing salaries,
then it isn't a business program, but if it checks salaries
then it is.
So how about, people, time or money?
I see, so a program that can read Victorian novels and answer
questions about them is a business program.
- Why do we program in COBOL? After all, COBOL is a pretty
specialized language that can barely run on a SUN 3/50.
You can obviously rewrite anything written in COBOL in C.
But C doesn't support pictures.
Good grief, you can write a library routine. Besides, who
makes enough money to really need commas.
And it isn't file I/O oriented.
You can use structures and call UNIX subroutines.
And it doesn't support decimal arithmetic.
Call a subroutine and call FASB and explain why the C way of
computing interest payments is better than their way.
And where's the report generator? I can get more done in 10
lines of RPG than 10 pages of hairy C code ....
Serves me right for arguing with a COBOL programmer..
Now that I've sundered every DP department to its philosphical roots,
they'll all start hacking 68020 machine code and deactivate my cash
machine card.
---- Sound familiar? ----
I'll make the argument that business people use languages like COBOL
(or 1-2-3) and AI people use languages like LISP (or LOOPS) because
they make it easier to write down the concepts they are dealing so a
computer (or other programmer) can understand them. You don't write a
loop to slide the dollar sign next to amount payable and you don't
shift bits around to see if an item is a fixnum or a rational.
Business programmers write programs that deal with the structures of
business. This means they deal with people, resources, time and money
or whatever else it takes to keep the business running.
AI programmers write programs that deal with the structures of
knowledge and reasoning. This means they deal with ontologies,
relationships, searching, recognition and transformation or whatever
else it takes to make a computer perform the tasks that are associated
with human intelligence.
I won't try and define "intelligence" here and I won't try and define
"business" here.
Seth
---- Letter Foot ----
------------------------------
Date: 18 Apr 88 16:52:09 GMT
From: govett@avsd.uucp (David Govett)
Subject: Re: AIList V6 #67 - Future of AI
>
> - Watch those Pony Express arguments. Remember, the Pony Express was
> a temporary hack. It ran for a bit under two years before it was
> replaced by the transcontinental railroad.
>
Not true. The PE was obviated by the transcontinental telegraph in
1861, I believe. The transcontinental RR wasn't completed until
May 1869.
------------------------------
Date: 18 Apr 88 1745 PDT
From: John McCarthy <JMC@SAIL.Stanford.EDU>
Subject: re: AIList V6 #72 - Queries
[In reply to message sent Sun 17 Apr 1988 23:35-PDT.]
McCarthy, John and P.J. Hayes (1969): ``Some Philosophical Problems from
the Standpoint of Artificial Intelligence'', in D. Michie (ed), Machine
Intelligence 4, American Elsevier, New York, NY discusses the problem of
free will for machines. I never got any reaction to that discussion,
pro or con, in the 19 years since it was published and would be grateful
for some.
------------------------------
Date: 18 Apr 88 19:32:00 GMT
From: channic@m.cs.uiuc.edu
Subject: Re: Free Will
I can't justify the proposition that scientific endeavors grouped
under the name "AI" SHOULD NOT IGNORE issues of free wil, mind-brain,
other minds, etc. If these issues are ignored, however, I would
strongly oppose the use of "intelligence" as being descriptive
of the work. Is it fair to claim work in that direction when
fundamental issues regarding such a goal are unresolved (if not
unresolvable)? If this is the name of the field, shouldn't the
field at least be able to define what it is working towards?
I personally cannot talk about intelligence without concepts such
as mind, thoughts, free will, consciousness, etc. If we, as AI
researchers make no progress whatsoever in clarifying these issues,
then we should at least be honest with ourselves and society, and find a
new title for our efforts. Actually the slight modification,
"Not Really Intelligence" would be more than suitable.
Tom Channic
Dept. of CS
Univ. of Illinois
channic@uiucdcs.uiuc.edu
{ihnp4|decvax}!pur-ee!uiucdcs!channic
------------------------------
Date: Mon, 18 Apr 88 20:09:47 EST
From: carole hafner <hafner%corwin.ccs.northeastern.edu@RELAY.CS.NET>
Subject: Future of AI
In article <1134@its63b.ed.ac.uk> gvw@its63b.ed.ac.uk (G Wilson) writes:
>I think AI can be summed up by Terry Winograd's defection. His
>SHRDLU program is still quoted in *every* AI textbook (at least all
>the ones I've seen), but he is no longer a believer in the AI
>research programme (see "Understanding Computers and Cognition",
>by Winograd and Flores).
>
I'm glad to see "Understanding Computers and Cognition" mentioned in this
discussion. It includes a lengthy section listing all the "justifications"
for AI, and then refutes them one by one. Conspicuously absent from this
list is "curiosity".
I think AI is the expression of the species' curiosity about this new artifact
(the computer) that it has invented. We want to find out what else can it do,
how smart can we make it, can we find a way to make it improve itself? Of
course, we have to pretend that we have socially relevant goals like helping
people or national defense in order to get the money to pursue these inquiries.
And sometimes the two (curiosity and socially relevant goals) are compatible,
since we need some tasks to focus our attention and test our theories.
Is this heresy? Or merely stating the obvious?
--Carole Hafner
------------------------------
End of AIList Digest
********************
