From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!yale.edu!jvnc.net!netnews.upenn.edu!libra.wistar.upenn.edu Thu Dec 26 23:57:13 EST 1991
Article 2281 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!yale.edu!jvnc.net!netnews.upenn.edu!libra.wistar.upenn.edu
>From: weemba@libra.wistar.upenn.edu (Matthew P Wiener)
Newsgroups: comp.ai.philosophy
Subject: Occam's barber had many customers
Message-ID: <60758@netnews.upenn.edu>
Date: 19 Dec 91 17:26:21 GMT
References: <60551@netnews.upenn.edu> <334@tdatirv.UUCP>
Sender: news@netnews.upenn.edu
Reply-To: weemba@libra.wistar.upenn.edu (Matthew P Wiener)
Organization: The Wistar Institute of Anatomy and Biology
Lines: 122
Nntp-Posting-Host: libra.wistar.upenn.edu
In-reply-to: sarima@tdatirv.UUCP (Stanley Friesen)

We seem to be agreeing more than disagreeing, so I'm going to cut this
reply down to some kind of minimum.  Also, I'm breaking it up into two
articles.  Sort of like ameoba fission, right?

In article <334@tdatirv.UUCP>, sarima@tdatirv (Stanley Friesen) writes:
>There is now a great deal of NN research that is aimed *directly* at modeling
>real brain operations.  This is why I put the "verified in BNS's" clause in
>the original.  I am requiring a certain amount of back-checking.

But how about my murkiness challenge?  To summarize, there are numerous
computational techniques inspired by nature: simulated annealing, genetic
algorithms, neural networks, and others.  When such algorithms are applied
to problems close in form to their initial inspiration, it is a priori
unclear whether any resulting success is due to the power of the method
or the accuracy of the model.  Simulated annealing of TSP is obviously
not real-life thermodynamics, but deciding this for SA of a protein
structure with a partially artificial energy is unclear.

In other words, NNs are too powerful to let us decide.

>			       But I suspect there may be more basic cellular
>factors at work than you may think.  For instance, I suspect that neurons use
>the rather slow expedient of axonal transport because living cells cannot
>easily manufacture wires with periodic voltage boosters.  Thus, I doubt if
>replacing the axons with boosted wires would be operationally relevant.

(I'm puzzled why you think I believe otherwise here.)

I'm reading Ira B Black INFORMATION IN THE BRAIN: A MOLECULAR PERSPECTIVE.
His thesis is closer in spirit to my X-cell-ent theory than any neural
wiring model.  He does not call on the immune system's known cognition,
but seriously proposes--with experimental support--that biochemistry is
the physical substrate for various mental processes/states.

See Black, Adler, Dreyfus, Friedman, LaGamma, Roach "Biochemistry of
Information Storage in the Nervous System" SCIENCE v236 p1263 (1987)
for an introduction.

>Bose-Einstein may well prove to be relevant to thought, but at present there
>is no evidence for it.  And since this is an additional mechanism, that is not
>observable in small groups of biological neurons, I will not postulate it until
>there is more evidence for it.

The only evidence so far is that it exists in "biological systems", and
that "biological systems" can exploit it.  Absolutely no evidence exists
to connect it with mind.

>O.K. 4) Everything that is not relevant to mental processes acts as an 
>        independent variable, to be explained by a seperate theory.

I believe that observation of the real world is relevant to certain
mental processes.

>In short, I am far from convinced that the other stuff Penrose wants to explain
>is even relevant to the operations of minds.

>He can explain it all he wants.  But he needs to *demonstrate* relevance
>if he wants me to treat his hypotheses as anything except philosophical
>vaporware.

Let me give Penrose's argument in a nutshell, ignoring all the cutesypie
bogosities that made his book so hopelessly annoying: QM is known to be
incomplete--it must merge someday with GR.  The result will be the end of
the line in ultimate physical theory.  In particular, it won't be subject
to philosophical divisiveness, so it will give us a clear cut account of
what observation is.  And without this theory, we won't ever really under
observation.

>|  But neither assumptions give such a theory, so I claim it is
>|premature to appeal to Occam.  3* already gives a link into the quantum
>|world, and so when we explain mind_and_world 1-2-3* plus 4-5 ends up being
>|"simpler" than 1-2-3 plus 4-5-6-7.
>
>O.K. But *why* do we *need* to link into the quantum world?
>I do not see it as at all necessary or relevant, as far as cognition
>or consciousness is concerned.
>
>*That* is the additional assumption that I do not accept.

Physicists connected consciousness with quantum mechanics over sixty
decades ago.  The debate has never ended.  Your 4 *assumes* one side
of this debate.  Penrose *assumes* the other side.

It is not a question of an additional assumption.  Penrose makes his
assumptions in roughly the same Occam-like approach that you do--but
he is working in a different constellation of things to explain.  You
*have* made an additional assumption--namely that QM follows certain
interpretations--but because this assumption lies outside the direct
focus of the questions you want to investigate, you can't see or care
one way or the other what extra philosophical baggage you have.

>You must show some observable cognitive processes that are *inconsistant*
>with a non-quantum explanation before such a link becomes necessary.

You are overstating the case.  A quantum explanation may simply be simpler
and/or superior.  QM is certainly more general.

Consider ambiguous word resolution.  There is evidence that all the
meanings are activated in parallel, and yet we only observe one at a
time.  I presume something similar happens with ambiguous figures.

The simplest interpretation is, quite simply, that there is a quantum
observable corresponding to "meaning", and by means of wave function
collapse, one eigenvalue--an unambiguous meaning--is observed.

A neural net model would have to couple with evolutionary dictates
that say fortune favors the unambiguous.  Entirely plausible, but
definitely more complicated.

>I find it interesting that the NN model of the olfactory cortex that has been
>developed shows the same EEG pattern as the living olfactory cortex.

I posted a separate article "Linear -> Non-linear theory bifurcation",
explaining how this depends crucially on a linearity in the bulb.  The
experimental agreement does not let you distiguish between internal and
external models of cognition.

If NNs were the key to cognition, I would expect that evolution would
keep that bulbar linearity in later developed neural groups.  Since it
didn't, I suspect that NNs are just a part of cognition.
-- 
-Matthew P Wiener (weemba@libra.wistar.upenn.edu)


