From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!usc!wupost!darwin.sura.net!mojo.eng.umd.edu!mimsy!harwood Wed Dec 18 16:02:49 EST 1991
Article 2248 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!usc!wupost!darwin.sura.net!mojo.eng.umd.edu!mimsy!harwood
>From: harwood@umiacs.umd.edu (David Harwood)
Newsgroups: comp.ai.philosophy
Subject: Re: Scaled up slug brains
Message-ID: <45115@mimsy.umd.edu>
Date: 18 Dec 91 18:46:53 GMT
References: <12773@pitt.UUCP> <45090@mimsy.umd.edu> <12781@pitt.UUCP>
Sender: news@mimsy.umd.edu
Organization: UMIACS, University of Maryland, College Park, MD 20742
Lines: 89

In article <12781@pitt.UUCP> geb@dsl.pitt.edu (gordon e. banks) writes:
>In article <45090@mimsy.umd.edu> harwood@umiacs.umd.edu (David Harwood) writes:
>
>>Call up NIH's center on stroke disorders, or McLean research institute
>>in Boston, or NYU Medical School's neuroscience project involving
>>special NMR imaging. (If you excise 60% of your left hemisphere,
>>which is involved in language processing (among other things),
>>except for areas B and W, you will not "proceed" as normal except
>>to post to this newsgroup.
>
>Will you get real?  Who said that the patient would proceed as normal?
>How in the world would you ever tell what B & W were doing if they
>are totally cut off from the rest of the brain?  No one said B & W
>were the totality of language processing.  But small lesions there
>can wipe out language and small lesions elsewhere cannot.  What is
>your theory of what B & W are doing?
\\\\\\\\\\\\\\\\\
	I'm serious. You cannot believe me, so call up experts at these
places - ask for reprints. You don't have to speculate about this -
find out; the answers are not what you seem to think, though.
	You are simply wrong - it is not a matter of plentiful "loose"
connections all over the brain. As I pointed out in the first place,
the neuro-linguistic activity occupies 80% of the left hemisphere,
not localized regions only (although these are necessary, I agree).
But the hemispheres are densely connected in normal brains, however
the right hemisphere only shows some lesser word-meaning related
activity -there is other evidence for this too. (A speculation might
be that the verbal processing of the right brain is "commutative"
-eg and/or gating or NN-ing of morphemic representation, while the 
left is non-commutative, non-associative, eg for recursive syntactic-
like patterns, or for causal orders, or kinaesthetic sequences, etc.
Who knows - but they are pretty clearly different hemispheres in humans,
but not in apes. 
	You ask for a speculation about what the other 60% besides B 
and W, etc are doing. Again and again, you edited out what I said, from
your mind and from your replies. I answered this. This other 60% provides
information for interpreting verbal expressions: if you "ablate," as you
said, this other 60%, you will not be able to make appropriate sense of
verbal expressions (even if you can repeat them). The memory and context
of experience is lost. (Call B and W parsers of input and output if you
like, although this is too simple no doubt. Anyway, when Penfield stimulated
cortical points outside these regions, people could verbally describe
what they remembered. To say this has nothing to do with language processing
is absurd.
	So we are not talking about "loose" connections in the brain
- there aren't any anyway (in a mature brain), since neurons are activated
in many modes of information processing. Most cortical cells do not
have identifiable specialized functions (as if they were characteristic
functions - for "grandma" or something). Their information is highly
abstract (5% of cells apparently even "count" up to small numbers of 
events of types! This in mammals. Most NNs can't count at all (although
you can modify them with recurrent links so they do).
	That is why I referred to the other discovery - that even
cortical motor cells were surprisingly activated by spoken words-
they process some sort of information- who knows what - related to
words. This is surprising because these giant cells have vastly rich
connections, and are "conservative" as you like to say, and have high
thresholds - it takes more than a "loose" connection to make them fire.


>Apes do not have the amount of neural hardware devoted to language
>that we do.  So what?  Who is disputing this?  But if you chew on
>their banana, it will activate large areas of the brain.  All this
>proves is that there are plentiful interconnections, nothing more.
\\\\\\\\\\\\\\\\
	If you talk to an ape, not very much will go on in his brain
- unless you yell at him, or steal his banana. It is wiser to do this
using a hi-fi system, for scientific and safety reasons.
	What you will see is not what you see in humans - you will 
see that both hemispheres are zero-or-more activated, depending on
whether you recite your favorite slogan, or whether you yell at the
ape. (If you recite you favorite slogan at Mr Z-, he will act as if you
are chewing on his banana, but we always have outliers in the data.)
	It is called the scientific method - it shows that ape brains
are very different from our own, despite superficial inspection. (By the
way, exactly how do you inspect all these neural circuits you have been
talking so sagely about, so that you know how "generic" they are among
species. You can only see a random disconnected 1% by conventional 
staining (thank God), and inter-neuron tracing techniques based on
labeling transportable metabolic products only show fragments of circuits.
	(BTW, do you realize that it takes an electron microscope to resolve
even an "impression" of a synaptic vesicle or gap? I wonder what the gaps
are on those problematic designs for computer chips.)
	Just curious about your superior methodologies, which warrant
such confidence about neuron morphology and function.



	


