Xref: utzoo comp.ai.philosophy:792 sci.psychology:4691
Newsgroups: comp.ai.philosophy,sci.psychology
Path: utzoo!utgpu!watserv1!ssingh
From: ssingh@watserv1.waterloo.edu (Sneaky Sanj ;-)
Subject: Re: Perceptron limitations...
Message-ID: <1991Apr2.214606.16223@watserv1.waterloo.edu>
Summary: WARNING: LONG... AND LONG-WINDED!!!
Organization: University of Waterloo
References: <1991Apr2.092041.9391@watserv1.waterloo.edu> <1991Apr2.145525.11793@watdragon.waterloo.edu> <1991Apr2.182825.4500@grebyn.com>
Date: Tue, 2 Apr 1991 21:46:06 GMT
Lines: 76

In article <1991Apr2.182825.4500@grebyn.com> fi@grebyn.com (Fiona Oceanstar) writes:
>Cameron Shelley writes:
>>Not only do our brains contain more than some minimum of
>>neural cells, but the cells come in many kinds and similar ones tend to
>>group themselves together.  The groups then tend to take on different
>>functions.  This kind of diversity is apparently part of what makes
>>`mind' possible.  Since the human brain is our best (understood) example
>>for mind, the factor of morphological diversity should at least be taken
>>into account.

Physical implementation aside, the bottom line is that they are still
devices that can be modelled by finite automata (or is this the
continuous vs discrete argument again ;-). The question still begs as
to whether or not if you string enough of them together the ability
for more complex computation arises that is not present in less complex
networks in any form, and is irreducible to any one of the elements.

With my convoluted understanding of neural nets_Perceptrons_ is the
only book that attempts to address this, and I was just pondering the 
notion that language is possible in humans because the capacity for
abstraction that underlies language can only be implemented by a
sufficiently complex brain. And do Minsky's results have any relevance.

>[...]  And I do agree with Cameron: models that view
>the brain as homogeneous, are hard for me to make heads or tails of--because
>the brain is so highly structured, so complex in three dimensions.

I agree that the brain is highly structured, but it is bad to immediately
trash models that abstract the brain as homogeneous. To model the brain
in this way allows for a generality that dwelling on the connectivity of
the hippocampus does not. Remember that we are dealing with a highly
refined and highly tweaked information processor. It makes sense from
an evolutionary standpoint to have a hard-wired link as outlined below.

stimulus -> iconic mem -> STM -> hippocampus -> LTM.

Presumably, links between iconic memory and STM are hardwired into place.
Via specialized structures that take away from the homogeneity of the brain.
Why? I would guess so that a high enough informational bandwidth can be
achieved to process information in real-time. If not, we might be a repast
for a mean sabre-toothed tiger! :-) Mind you, the specialized structure
of the hippocampus serves a different function, but it may need to
be "optimized" in an analagous fashion.

So anyhow, I maintain that homogenous models are good and convenient for
simulation and theoretical results. Domain-specific optimization is best
left to field-testing. And homogeneous models areable to achieve
functional equivalence to more specialized models, even if real-world
implementations are not as effective. 
 
>[neat story about monkeys deleted]

>And they say only humans have language.
>
>							--Fiona O.

Your example seemed to imply that it was very much like classical
conditioning. Where a certain stimulus led to a certain response.

This is not language. These monkeys most likely do not have the
ability to communicate the symbol for "eagle" or "snake" without
being convinced of seeing such a thing; ie. stimulus -> response.

That's the difference. I can type "snake" and you know what I'm
talking abou. I don't have to physically bring you a snake or 
appeal to sense datums to communicate the concept of a snake. I doubt
that those monkeys could do that.

Sanjay Singh never existed... There was only...
Ice.

-- 
"No one had the guts... until now!"  
$anjay $ingh     Fire & "Ice"     ssingh@watserv1.[u]waterloo.{edu|cdn}/[ca]
ROBOTRON Hi-Score: 20 Million Points | A new level of (in)human throughput...
!blade_runner!terminator!terminator_II_judgement_day!watmath!watserv1!ssingh!