Re: NEURO: Advanced Neurons?

Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Tue, 18 Mar 1997 09:44:36 +0100 (MET)


On Mon, 17 Mar 1997, John K Clark wrote:
> [retinal extrapolations]
> There is nothing trivial about it, the density of neurons in the retina is
> 100 times greater than any other part of the brain, Moravec used it because

Though I haven't heard about such densities, it strenghtens the case of
retina being very special. Extrapolations based on the retina
(extrapolations, not speculations) will hold merely for the retina, and
not anywhere else.

(Apropos Hans, he now has moved entirely into the realm of religion with
his recent writings, even way beyond Tipler, whose Omega point theory at
least makes several falsifyable predictions. This not to belittle the man,
merely I don't like the direction he is currently heading to. It tends to
make one passive. A belief system can be a dangerous thing).

> it's the most studied part of the brain. I think it's a quit reasonable thing

Sure, it's _comparatively_ simple and easily accessible.

> to do. Moravec assumed 10 bits per synapse, also quite reasonable at the time
> but we now know that's way too high.

I won't comment on that. I mean I already did, but I won't.

> [...]
> >minimal threshold may lie quite high for such complex objects as a
> >mammal brain.
>
> No need to say "may", you will definitely need to do a hell of a lot of
> computations to emulate any mammal's brain, and there is no reason why a very
> small object couldn't perform a hell of a lot of computations.

A bee's, or a fly's brain is sure a very small object (1 mm^3, or so),
and it sure does an awful lot of computations to survive in a complex,
unfriendly environment (hmm, this makes me want to rush out instantly, and
read up on insect neuroarchitecture). The question is: how much smaller
can the system get? How much faster? At equivalent speed, what would its
power requirements be? Intuitively, or, as an educated guess, one, two
orders of magnitude, no more.
> [how many persons can fit in a waterglass?]
>
> Well of course it's Science Fiction, nobody has done it yet, and 10^9 is

The main difference between "fantasy" and "science fiction" is imo "science".

> probably closer to the truth.

I think you are way off the correct numbers, but I won't repeat all the
arguments why.

> >not even very good science fiction.
>
> For me good Science Fiction extends current ideas to the absolute limit
> without breaking the laws of Physics. I think a sugar cube sized super brain

"without breaking the laws of Physics". Yes. Trouble is, basic laws do
not allow us to talk about constraints of complex objects. We know
(assuming no femtotech) atoms to be the smallest utilizable structures,
(not even atoms but clusters of them) to define the lower horizon of
implementability, and we can observe biological information processing
systems. We know electronical systems to be much faster than mechanical
ones, and we know maximal switching speeds in a given volume dictated by
bond photodissotiation energies. We don't know whether we will have
drextech, with its diamondoid matrix and fractal cooling channels,
allowing it to operate at astronomic power densties, without turning
into lava instantly.

There is still a whole lot what we do not know yet. So I prefer to take a
conservative stance (in mundane life most people think I am a lunatic,
but here I'm actually a conservative. Amazing), which is less embarassing
on the long run. But then people forget.

> is very good Science Fiction.

Well, it certainly makes a jolly good read.

> >Darwin doesn't operate on individuals, but on populations.
>
> If the unit of natural selection is the population then what constitutes a

Perhaps I should have stated that differently. A population of
imperferfectly autoreplicating entities in a limited-resource environment
is a necessary and sufficient condition for darwinian evolution to emerge.
Better?

Meaning, no population, no evolution. No more, no less.

> population? Are The Extropians a population? How about a nation, or the
> Human Race, or primates or mammals or vertebrates or animals or life in
> general?
>
> Darwin doesn't operate on individuals, but on genes, the basic unit of
> reproduction.

A gene does not reproduce outside of its wet context, the animal cell.
Besides, darwinian evolution is not limited to systems having genes (or
bits, or whatever), just inheritable, mutable properties which have impact
on system's fitness.

ciao,
'gene