IUBio Biosequences .. Software .. Molbio soft .. Network News .. FTP

machine brains

Michael Edelman mje at mich.com
Tue Feb 16 13:59:11 EST 1999



Ray Scanlon wrote:

> What do we wish to simulate and at what level? Are we talking, for instance,
> of the brain of rat? This is a mammalian brain, it has all the parts of a
> man's brain, just not so many neurons? What is the unit of our simulation?
> The neuron?
>
> The neuron is a large assortment of neurons, I have no idea how many. I
> would be happy if someone would give me a guess or a citation as to how many
> in an average neuron. Some would say that until we know the story of every
> molecule we cannot simulate a neuron. This is hogwash.

We can simulate a neuron, but we need to know more about how neurons
interconnect. At one time we thought all neronal connections were
axon-to-dendrite. Then we found axon-axonal, electrical, even dendrite-dendrite
connections. Without an understanding of all those mechanisms we can't
completely model a single neuron.

Moving up to small networks, it's only recently that the role of nitrous oxide
has been revelealed as a diffuse, local neurotransmitter; unless you model that
as well, you don't have a full model of the physical level of the brain. And
there's no doubt more to be discovered.

> What we simulate is what interests us. We are interested in the neuron as an
> electromechanical device that influences other neurons. The myriad of
> molecular activities that have to do with the maintenance of a living cell
> are not involved.

Ah, but they are. Axonal transport regulates the rate at which an axon can
produce vesicles of transmitter. Availability of calcium can regulate firing
activity. All these can be influenced by metabolism, mitochondrial activity and
so on. The cell wall is pivotal in transmission of an AP to an axon. We need at
least a functional understanding of the relationships.

There's a danger here in defining the question down to the point where it's
simple, but uninteresting. Supposing you assemble 200 million of your synthetic
neurons into a mass and start generating S-R pairs. Do you have a mind, or just
an interesting programable array?

> The endless molecular activity of a synapse can be lumped
> together with a few parameters. The synapse is strengthened or it is
> weakened. The synapse disappears. there is axonal growth and new synapses
> appear. All these are subject to a time dependence (from birth to death of
> the organism). A synaptic event involves a local change in the potential
> difference between the interior and the exterior.
>
> This is not very much as simulation goes.

It's an incomplete picture. There's a lot more going on than simple
axonal/dedritic transmission.

> There is no problem as long as we keep the soul (mind) out of it.

Hmm. I think I'd argue this point. What are we trying, in the end, to explain?
Mind. If our brain model doesn't actually make a stab at explaining mind, what
good is it? It's just another hypothetical model of nothing in particular, like
so many other AI models.

> The brain
> is simple and its activities are simple. We are overwhelmed by the number of
> neurons, by the number of nuclei, by the number of tracts, but if we lump
> the trees together we shall see the forest.

Of course, I think the forest is Mind.

--
Michael Edelman     http://www.mich.com/~mje
Telescope guide:    http://www.mich.com/~mje/scope.html
Folding Kayaks:     http://www.mich.com/~mje/kayak.html





More information about the Neur-sci mailing list

Send comments to us at biosci-help [At] net.bio.net