Copy Link
Add to Bookmark
Report
AIList Digest Volume 5 Issue 173
AIList Digest Thursday, 9 Jul 1987 Volume 5 : Issue 173
Today's Topics:
Humor - Symbol Grounding References,
Theory - Fuzzy Categories,
Policy - Symbol-Grounding Metadiscussion
----------------------------------------------------------------------
Date: 7-JUL-1987 15:50:42
From: UBACW59%cu.bbk.ac.uk@Cs.Ucl.AC.UK
Subject: References Required.
Does anyone have any pointers to the "symbol grounding problem" or some
such area? Searches in the literature have proved fruitless.
The Joka.
------------------------------
Date: 7 Jul 1987 11:00-EDT
From: Spencer.Star@h.cs.cmu.edu
Subject: Re: AIList Digest V5 #169
> ...a penguin is not a bird of degree...
The point of view that a bird IS a bird, and a rose IS a rose, has
limited usefulness. If the question that we are trying to answer is
seen as how a person will classify a penguin after having seen one for
the first time, I think the answer is clear. A large number of people
would not classify a penguin as a bird. A program would likely be more
successful at imitating a human response if it based its response on
the features of the human answering the query as well as the features
of the concept it was trying to recognize. Whether a penguin is a bird
then becomes quite dependent on context as well as a simple relation
between features and classes.
------------------------------
Date: 8 Jul 87 16:08:27 GMT
From: sunybcs!dmark@ames.arpa (David M. Mark)
Subject: Re: The symbol grounding problem: "Fuzzy" categories?
In article <974@mind.UUCP> harnad@mind.UUCP (Stevan Harnad) writes:
>
>
>In Article 185 of comp.cog-eng sher@rochester.arpa (David Sher) of U of
>Rochester, CS Dept, Rochester, NY responded as follows to my claim that
>"Most of our object categories are indeed all-or-none, not graded. A penguin
>is not a bird as a matter of degree. It's a bird, period." --
>
>> Personally I have trouble imagining how to test such a claim...
>
>Try sampling concrete nouns in a dictionary.
Well, a dictionary may not always be a good authority fro this sort of
thing. Last semester I led a graduate Geography seminar on the topic:
"What is a map?" If you check out dictionaries, the definitions seem
unambiguous, non-fuzzy, concrete. Even the question may seem foolish, since
"map" probably is a "basic-level" object/concept. However, we conducted
a number of experiments and found many ambiguous stimuli near the boundary
of the concept "map". Air photos and satellite images are an excellent
example: they fit the dictionary definition, and some people feel very
strongly that they *are* maps, others sharply reject that claim, etc.
Museum floor plans, topographic cross-profiles, digital cartographic
data files on tape, verbal driving directions for navigation, etc., are
just some examples of the ambiguous ("fuzzy"?) boundary of the concept
to which the English word "map" correctly applies. I strongly suspect
that "map" is not unique in this regard!
------------------------------
Date: Mon 6 Jul 87 16:18:12-PDT
From: PAT <HAYES@SPAR-20.ARPA>
Subject: Re: AIList Digest V5 #170
Talk about walking into a minefield, but here goes. Concerning the Harnad
grounding problem. This is lovely stuff, and I save every word for later
reading, but it does seem recently to have gone from interesting discussions
and arguments to a rather repetitive grinding over the main points again and
again. THe result is that Stevan is reduced to repeating himself and
reiterating his points in the face of what must seem to him to be increasing
stubbornness. I seem to be seeing more and more phrases like '..as I have
emphasised earlier..'. All of us who teach are familiar with the syndrome
where the 35th occurrence of the same error makes us more exasperated than the
first one did.
Let me suggest that perhaps nothing much new is being said
in these discussions any more, and certainly no-one is saying anything which
is going to cause Stevan to change any of his positions. Perhaps the right
thing to do is for people to send their comments directly to Harnad, and for
him to send us the selections which HE considers worth public airing, together
with his responses. That way we will be spared reading all this stuff which
is, apparently, of such low intellectual caliber, and Laws will have an easier
time, and public feelings will not get to the point which produces letters
like David Harwood's.
Just an idea.
Pat Hayes
------------------------------
Date: Mon, 6 Jul 87 16:56:06 PDT
From: cottrell%ics@sdcsvax.ucsd.edu (Gary Cottrell)
Subject: Automatic newsgroup creation to reduce aggravation
How about some software to automatically create newsgroups after a certain
amount of traffic with the same subject line? And an appropriate expiration
of the newsgroup after traffic dies down? Then people could decide to add
the newsgroup or not. E.g., comp.ai.symbol.grounding.. It doesn't even sound
hard enough to be called AI! I am a net.news.software.innocent, however.
gary cottrell
------------------------------
Date: Mon, 06 Jul 87 22:03:51 EST
From: Tim Daciuk <ACAD8023%RYERSON.BITNET@wiscvm.wisc.edu>
Subject: Symbol Grounding Problem
Having read the recent "discussion" regarding the Symbol Grounding Problem,
I would have to suggest that I tend to agree with Mr. Harwood. Though the
discussion which has taken place on this subject was interesting, it has
become, at least to me, tedious and boring. In addition, I think that any-
one joining AI-List at this point would find this topic almost impossible
to follow, due to the number of references to previous editions of the
journal, and due to the highly interactive mode which this discussion has
assumed. I do not think that a separate discussion should be started,
however, I would suggest that future Symbol Grounding Problem entries be
sorted to the bottom of the list. This would allow the list to continue
in publishing this important part of AI, and would allow those of us who
no longer have the stamina to ponder the implications of blue, green, blue-
green, etc., to quit at an appropriate time.
Would sorting the list with Symbol Grounding coming at the bottom be very
difficult Ken?
Tim Daciuk
Ryerson Polytechnical Institute
Toronto Ontario
Canada
[That's essentially what I've been doing, although lengthy conference
announcements sometimes get sorted even lower. I was holding all of
the symbol grounding discussion for the weekend, although that did
create some synchronization problems between messages sent to my
Arpanet mailbox and replies that went directly to Usenet. I have
usually published symbol grounding issues in separate digest issues,
making them easier to skip (or save). Usenet readers don't get the
benefit of that sorting, of course (but make up for it by eliminating
the digesting delay). Sorting to the true "bottom" of an infinite
discussion stream would seem a little extreme. -- KIL]
------------------------------
Date: 6 Jul 87 21:38:29 GMT
From: harwood@cvl.umd.edu (David Harwood)
Subject: An apology for being overly sarcastic
I want to apologize for being overly sarcastic with Mr. Harnad.
Although I consider my complaint about his postings to be justified, I am
sorry about my overly-sarcastic manner. For the record, this apology was
my own idea, not involving discussion with others. I simply felt fairly
guilty about my irritable responses. (Actually, it is only recently that
I've had a chance to read this newsgroup; it has been suggested that I
read the moderated newsgroup instead - without posting of course!)
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
Briefly responding to a posted reply by B.I. Olasov, also to
correspondence by email from D. Stampe (for different reasons):
In article <1071@bloom-beacon.MIT.EDU> bolasov@aphrodite.UUCP
(Benjamin I Olasov) writes:
[...]
>Some of the most challenging and interesting problems of AI are philosophical
>in nature. I frankly don't see why this fact should disturb anyone.
>
>Perhaps if more of us pursued our theoretical models with comparable rigor
>to that with which Mr. Harnad pursues his, the balance of topics represented
>on comp.ai might shift .....
As I tried to make clear - supplying fairly clear examples of his
posting style - it is definitely not Mr. Harnad's particular philosophy or
theoretical proclivity which irritated me - it was his manner of discussion
I was complaining about. (Just as others complained about my sarcasm, more
than the content of my complaint.) Among other things, for example, I
scarcely consider his arguments to be what you say "rigorous." Some of the
discussants themselves have complained, albeit politely, about his somewhat
idiosyncratic usage of terminology (among other things).
So you are mistaken in your suggestion about my complaint. Rigorous
use of very complex and abstract concepts is commonplace in many branches
of computer science, eg. semantic specification of languages executed by
parallel systems. The level of abstraction and rigor is not at all less
than in any area of inquiry, including philosophy or cognitive psychology.
On the other hand, I fully agree that both philosophy and psychology have
very important and relevant contributions to what is called "artificial
intelligence," although it seems to me that too much of the purported
interdisciplinary discussion is polemical and political rather than really
constructive. And I would add that much, even most, of AI's recent "advance"
has been nonsensical propaganda for funding, and devoid of theoretical
foundation.
Also, I would add that Mr. Harnad - what is clear by his
postings - is perhaps only superficially familiar with what are real
advances in symbolical "AI", eg development of very powerful systems for
automatic deduction, which have practical importance for all of "AI"
as well as have rigorous foundations. These surely are not entirely
founded on theories of human psychology or on speculative philosophy,
and probably should not be, since we would like to consider computing
machines which do some things according to specification, and better
than we do.
I realize very well that some areas of AI are very much
harder than others - computer vision comes to mind ;-) and it is
obvious to everyone concerned that we need both numerical and symbolical
algorithms and representations. (I will not get involved in discussing
what S.H. might mean by "symbolical", "analog", "invertible", and so
forth - I don't really know.)
I think it is also apparent that we might have yet
to consider some "connectionist" architectures and algorithms, which
perhaps do not admit any simple formal specification of input/output
relations. This would invite some philosophical speculation about
the adequacy of purely logical specification for development of
artificial intelligence. Conversely, we may already have sufficient
theoretical basis for 'creating' human-like artificial intelligence,
by functional simulation of neurons, although we do have the technology
(and moral sense I hope) to de-engineer a human brain. This will surely
happen in the distant future only depending on our technology and not
on major improvements in our theoretical understanding of neurons. The
situation might well be that we can recreate human intelligence which
we still largely cannot comprehend by formal specification. In part,
these means that psychology, theoretical "AI", even S.H.'s "Total
Turing Test" are loose ends as much as interdependent.
(As a religious person, I wonder about what this might mean -
I recall that an ancient interpretation of the Genesis story said that
when mankind ate of the fruit of the knowledge of good and evil - just
as the serpent claimed - mankind became endowed with a power like that
of God - that is, having the power to create and to destroy worlds. In
our times, our technology has surpassed our moral sensibilty - which
many computer sceientists say does not exist anyway. Of course, other
Jewish tradition has it that many worlds have already been destroyed
before this one. I'm not even sure that pursuit of "AI" technology
is such a good thing, if it contributes to our destruction or loss of
dignity. But who knows, except for God?)
Response to a reply by email:
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
Message-Id: <8707061548.AA11938@uhmanoa.ICS.HAWAII.EDU>
Date: Mon, 6 Jul 87 05:36:04-1000
From: seismo!scubed!sdcsvax!uhccux.UHCC.HAWAII.EDU!nosc!humu!stampe
(David Stampe)
To: harwood@cvl.umd.edu (David Harwood)
In-Reply-To: harwood@cvl.umd.edu's message of 5 Jul 87 21:48:28 GMT
Subject: Re: The symbol grounding problem - please start your own newsgroup
Status: R
You have now posted four messages to comp.ai containing nothing
but rude complaints about another's postings on symbol grounding.
They are not required reading, and they don't prevent you from
reading or posting on other topics. What you MAY NOT do is
disturb the newsgroup with irrelevant and loutish postings like
your last four. There are people who care about how University of
Maryland employees behave in public.
If I were you, I'd consider a public apology.
David Stampe, Univ. of Hawaii.
\\\\\\\\\\\\
I don't have any desire to prevent S.H. from posting,
as I have made clear. You are right that I should apologize for
being overly sarcastic. He deserved some of it, but I overdid
it.
I don't like your mention of my employment here - which
might be considered to be a threat, either to my employment or
to post things which you dislike, even sarcastic complaints. If
you did threaten me like this, you would have misjudged me, also
misjudged what would be my reaction.
In any case, you are right about the apology being due.
-David Harwood
------------------------------
Date: Tue, 7 Jul 87 07:33:49 edt
From: dg1v+@andrew.cmu.edu (David Greene)
Subject: handling the S.G.P issue
While some of the discussion has proven interesting, it is become burdensome
to sort through and rather recursive as arguments start focusing on what
prior arguments meant...
Perhaps a seperate bboard would be more appropriate. At the very least, Ken
Laws' suggestion that the arguments (and subject lines) be broken into
discrete categories seems to go a long way toward making this disscussion
palatable if not worthwhile.
Mr. Harnad might want to consider proposing a subject taxonomy prefaced with
"SGP".
David Greene
Carnegie Mellon
------------------------------
Date: Wed, 8 Jul 87 11:02:53 GMT
From: Caroline Knight <cdfk%hplb.csnet@RELAY.CS.NET>
Subject: Debating
As a so-far passive reader of the grounding problem debate via
AIList Digest I have at last been spurred to action:
For the proponents of a theory to be able to understand and discuss
the positive, the negative and the intersting aspects of it is a sign
of strength. For them to resort to personal name calling is not.
However I do have sympathy with those who have now started to put the
boot in. Especially with those who are tired of the language which
is frequently unclear and suspiciously polysyllabic.
A thought for those who honestly believe that an idea is wrong and the
holder of it would be better off without it:-
1. It is much easier to change one's mind and throw away useless ideas
if one has NOT been pushed to defend them tooth and nail.
2. Few ideas (or accepted theories) are completely correct. One can
gain more by simply acknowledging that an idea has flaws than by
trying to stretch it until it rips. Of course these anomolies might
trigger new ideas.
Caroline Knight
------------------------------
End of AIList Digest
********************
