X-Message-Number: 26862
Date: Sun, 21 Aug 2005 09:09:56 -0400
From: Thomas Donaldson <>
Subject: answers to Yvan Bozzonetti

For Yvan:

1. On the question of whether or not a parallel processor with N 
processors can be precisely simulated by one with M < N processors,
I have no references but can point out some obvious examples: suppose
that you a fleeing a tiger, and for that reason you have all your 
muscles and senses working SIMULTANEOUSLY on the task of fleeing
the tiger. A single processor could not do even part of this task
without failing to do its different parts simultaneously. 

Suppose each processor in the M processors was much faster than 
a single processor in the N processors. Yes, depending on what you
have to do, you could use such a system to simulate some of the
actions of the N processor brain, though it would be clear that
close observation would show that the N processors were working
simultaneously while the M processors weren't. Perhaps your machine
could be fast enough to flee tigers. Still, sooner or later 
something would come along which the N processors could handle
while the M processors could not: a barrage of cannon fire,
missles, and other things much faster than tigers.

2. The internet is physically spread out internationally. Nor, when
compared to formation of new connections in brains, does it work
very fast. Your first task here is to compress the internet into
something the size of a human brain; your second task is to make
it work faster. And the internet also requires special software to
work at all; this software would need to be modified so that
it worked more like a brain. (I'm sure that all the users  of 
the Internet would be delighted :-)).

The crucial point here is the telephone wires or radio waves from
satellites with which the Internet works. Since in many countries
everyone already has a telephone, the Internet can use that system
to make its connections. In a real brain, there may not be any 
connection between two neurons, so that it would need to be 
created, just like installation of a telephone (or setting up
a standardized radio connection amid all the other such connections).

3. Synapses are complex. It's a separate question whether or not
they work like parallel machines, with every single feature of
a synapse working at once. I don't know of any neuroscience
experiment which looks at that question: most work on synapses
deals with issues such as their responses to different transmitter
chemicals and what structures and chemicals produce those 
responses. A lot of complexity comes from the series of chemical
responses a synapse makes to a chemical input.

If you find parallelism in the operation of single synapses (I
don't mean complexity but parallelism) I would certainly be
interested. It would, of course, make our brains even more
parallel than they are now, and increase the problem I discussed
in 1). above.

As for creating a computer simulation of a neuron, that would
be interesting and valuable, even if we'll need to keep aware of
the boundaries of that simulation ie. the points at which it
fails. Point 1 above tells of one such point at which such
simulations can fail.

            Best wishes and long long life for all,

                Thomas Donaldson

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=26862