[net.misc] Architechture & AI

zrm (12/05/82)

A couple things about architechture & AI: The suggestion of connecting
itty-bitty processors and letting them run at random is very similar to
a spoof of AI research in "A Stress Analysis of A Strapless Evening Gown
& Other Essays for a Scientific Age" called "The Chaostron" or some
such. But that is on the right track.

Yesterday I attended a lecture about "linguistic chimps", one of the
lastest pop-science crazes. As it turns out, chimps are completely
incapable of syntax, not to mention other dubious aspects of these
experimenmts. The lecturer then went on to explore the differences
between chimp and human brain. There are remarkably few differences.
There *seems* to be no physical explaination for such a gross deficit in
language aquisition as a total lack of syntax. Indeed chimps aquire
other skills much faster than human childeren, and also aquire large
vocabularies. So why no syntax?

The key seems to be development time. Chimps are born more developed and
develope much faster than human childeren. Chimps are born with 60% of
adult brain-mass, and make up the 40% they dont have much faster than
human childeren, which are born with 24% of adult brain-mass and take
15-20 years to get up to full brain weight. Further, there is an
evolutionary artifact that manifests itself in a disorder that causes
premature puberty at about the same age as great apes mature. The rest
of childhood is a tacked-on kludge. Without all this unmade brain
constantly being subjected to stimuli as it develops, we would not be
the linguistic creatures we are.

Dolphins seem to be far more promising. Does anyone know about dolphin
studies? Unfortunatly this field is awash in charlatans.

Cheers,
Zig

alj (12/15/82)

This is a followup to the article referring to monkey's inability to deal
with syntax:

I believe that Schank would say that syntax is pretty much irrelevant to
human understanding of natural language.

Got that?

Compilers are neat, but we're neater still. . .

Andy Judkis
BTL Lincroft

wsp (12/16/82)

If, as Shank says, syntax is irrelevant to human language understanding,
what purpose does it serve?  Syntax is complex, takes many years to learn,
and differs markedly from one language to another (although there are
some deep correspondances).  I have always believed that evolution is pars-
imonious, i.e., vestigal and useless appendages tend to disappear.  Why 
would we retain something as obtrusive as syntax if it served no evolutionary
purpose?

Peter Benson
ittdcd-west
(619) 578-3080

debray (12/21/82)

I agree that syntax is not AS important to natural languages as it is to
programming languages (since, obviously, we can understand sentences even if
they are somewhat ungrammatical). However, I can't agree with Schank when he
calls syntax "irrelevant" to natural languages : it's obvious that very often,
small structural (syntactic) changes to a sentence will change its meaning
significantly. Try the following two sentences, where the only difference
between the first and the second is in the punctuation (and punctuation, as
we all know (I hope!), is syntax) :

	Mary loves a man who plays tennis, and runs.
			( => Mary runs ; Mary loves a guy who plays tennis.)

	Mary loves a man who plays tennis and runs.
			( => Mary loves a guy ; that guy plays tennis and runs.)
The difference between natural and programming languages in this regard is due,
I think, to the vast real-world database at our disposal, which is implicitly
referred to in our everyday use of natural language ; that information is
so helpful in helping us understand the meanings of sentences that syntactic
errors are not (necessarily) fatal.
Richard Montague's work on the relationship of natural language syntax and
semantics is a very interesting approach to the issue.

I agree with Andy, though, that we're much neater than compilers!



Saumya K. Debray
SUNY @ Stony Brook