X-Message-Number: 26078
Date: Sun, 24 Apr 2005 09:48:21 -0400
From: Thomas Donaldson <>
Subject: Some answers for Yvan

For Yvan Bozzonetti:

Even if you combine "only" 1000 neurons on a single chip, and have them
work by time-sharing, you'll still eventually encounter simultaneous
events on which your time-sharing brain will crash. It doesn't require
all neurons to time-share for such a thing to happen. I didn't even think
that you were planning to time-share your program for all neurons.

As for producing growth of neurons by means of software, your supposed
idea fails completely. First, consider a neuron, with its axon and its 
dendrites. When a neuron grows a new connection, this connection 
produces a REAL connection between two neurons. It shows its reality
because messages can be sent from the axon of one neuron to some of
the new dendrites of the other. To make such a connection clearly
requires physical changes to both neurons. The most software could
do is to provide a plan for how to put together the new connections.
Again, we get an even worse problem if we consider growth of entirely
new neurons (the ventricular zone in our brains is lined with cells
which can turn into neurons, migrating to their final site; the dentate
gyrus in our hippocampus also grows new neurons. Recently, I reported
in PERIASTRON that quite unlike many other mammals such as rats, 
most of the cells from OUR ventricular zone and that of monkeys DID NOT
end up in the olfactory cortex --- so for humans and some other animals,
not only do we grow new neurons, but they end up all over our brains).
Again, the very most software could do is to produce a plan for these
neurons; the entire process of growth would still have to take place
in a real brain.

Would you mind, then, explaining exactly how your brain will have
neurons which grow new connections, and even sometimes grow new 
working neurons? Even if you have things set up so that a chip
holds 1000 time-sharing neurons, growing new neurons will raise
problems which separate neurons would not: say that it wants to connect
to one of the neurons on a FULL chip. Then it must somehow do so
while keeping the connections of all 999 other neurons unchanged and
with the neuron it wants to connect to somehow growing one more
connection... or more if we're talking about dendrites.

I won't take back what I said about growing new neurons in software.
I'll also add a short comment: many people think that our brains
must be slow because all its neurons are slow. They forget that
our brains are highly parallel computers, and hence the speed of
a single neuron bears little relation to the speed of a brain. The
main problem with speed and brains comes from the simple fact that
we haven't evolved to solve the big problems that our fast computers
solve: we're very slow on big partial differential equations, for
instance... and many other presently valuable problems, too.

            Best wishes and long long life for all,

                   Thomas Donaldson

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=26078