X-Message-Number: 24060
Date: Fri, 7 May 2004 10:58:31 -0400
From: Thomas Donaldson <>
Subject: CryoNet #24054 - #24059

Hi for those interested in making brains:

Several points: 

1. Yes, I did not consider the possibility of connecting the neurons
   by radio or some other method which does not involve direct
   connection. You'll still have to have very many connections.

2. OF COURSE we can have messages passing through several nodes. We
   do that right now with what we call telephones. However there
   is still a loss in speed and a significant amount of switching
   devices and software to connect the number of neurons in a human
   brain. Furthermore, if our device grows new connections rather
   than connects to them by a phone line, it will reach those to
   which it's connected much faster. 

   The problem comes from the number of POSSIBLE connections such a 
   machine can have. For any single instance, it's true that it 
   will use only a small subset of those possible connections. However
   if we limit our machines to some arbitrarily chosen smaller
   subset, they will ALL be limited, which is hardly what we want.
   If on the other hand we give them the possibility of contacting
   all possible (or all reasonable) nodes, then they'll all be carrying
   around A LOT of unnecessary hardware. Hardly a good idea.
      
3. How important is speed here? I strongly suspect that natural 
   selection has basically adapted us to an "appropriate" thinking
   and mental processing (most of our mental processing is unconscious)
   speed. Either running faster starts to become more expensive, or
   it gives no special advantage in the lives we now lead. Yes, there
   are occasions in which (if we could) it would be good to react
   faster, but those reactions also involve physical movement and not
   just the speed with which we think.

   And we could get into a long and involved discussion about how much
   mental and physical speed we might get with a redesign. We'd probably
   have to miniaturize ourselves to start. However I won't get involved
   in such a discussion in this message. I'll just say that optimal
   speed is rarely maximum speed.

4. It's important to consider BRAINS here rather than neural nets. Among
   their other features, neurons usually have many more connections than
   the nodes in a neural net, and those connections behave in several
   distinct ways depending on just which connection you're looking at.
   There are directly electrical connections and also those which use
   a transmitting chemical, of which there are a fairly large variety
   of different transmitting chemicals. So those connections for brains
   don't just send a single YES or NO signal. And yes, our neurons do
   grow new connections and our brain makes new neurons, too. To see
   all this, just get hold of a diagram of a pyramidal neuron, and 
    remember that you're not looking at a generic neuron but one of
   a frequent, common, and important type but hardly universal.

   Not only that, but real neurons might be better represented as 
   nodes with a large set of branches, with one branch (the axon)
   going out from one end and also ending in branches.

   If you wish to represent such a system as a neural net, it will
   have to have more than one node for each single neuron, and multiple
   versions of each connection for the different types of messages
   that a neuron can send or receive.

5. Yes, my calculations weren't quite correct, but here is a better
   version --- and note that you still get a good-sized object. If
   we work with square connectors and want our node to be a box,
   then the box will be a bit more than 10 microns on a side. In terms
   of making the circuitry in your node nanosized, you'll have to
   forget that; your node will be too big. The area of the surface
   of a box is 6 * N^2 where the box is N X N X N. If the connectors
   are square in cross section and 1 nanometer on a side, then if we
   line them all up we get a length of one meter. We don't want to
   line them up, we want to collapse them into a square. The 
   square root of 1 billion (1 nanometer == 1 billionth of a meter)
   is ~~ 3 x 10^(-4.5) ~~ 10^(-4) ~~ 10 microns. This is a pretty
   good sized node for an electrical circuit --- especially remembering
   that it contains NO extra space.

Most of all, remember that brains are not neural nets, nor are they
any kind of computer yet made. And after all, just what is the objection
to growing new connections, anyway? 

                 Best wishes and long long life for all,

                    Thomas Donaldson

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=24060