Re: Edge.org: Jaron Lanier

From: Eugen Leitl (eugen@leitl.org)
Date: Tue Dec 02 2003 - 15:03:38 MST


On Tue, Dec 02, 2003 at 01:46:42PM -0500, Eliezer S. Yudkowsky wrote:

> Taken literally, this seems to me to contradict the second law of
> thermodynamics. A system that evolves in the presence of noise may

A computational system is driven by an entropy gradient, but it
can do whatever computations it damn pleases within a given energetic
envelope. There's no perpetuum mobile there.

Spiking models do more computation than continuum models; noisy spiking
systems do more still. There's something in spikes and noise that go
together.

> require the presence of noise to go on working. An algorithm that does

Sure, but that is not a handicap. Noise is cheap at nanoscale, in fact
you have trouble telling your signal from noise there. It is very simple
to tap analog noise in circuits, if one doesn't insist in cryptographic
whitening.

> exploration using a merely pseudo-random algorithm may degrade even

You can't tell the difference between RNG and PRNG even for trivial
sized bit assemblies.

> further if the pseudo-random steps are replaced by steps with heavy
> correlations. A system that evolves to tolerate noise will also,
> usefully, tolerate some other errors that look like noise, and will
> tolerate some useful mutations that introduce noisy side effects.

You're vortexing several layers simultaneously here. The noise the
neural system sees has nothing to do with genome mutations. I'm probably
misunderstanding you here.

> Evolution (a noisy design process) in the presence of environmental noise
> is one particular way to get flexibility and robustness in the face of
> design and environmental changes. But you cannot actually use noise to do
> computation any more than you can use thermal noise to turn a ratchet

I can't pull up the relevant references right now, but, yes unreliable
synapses and system noise do seem to go together.

> wheel. Mystification of noise is part of a general mystification of
> neurobiology.

Neurobiology doesn't need extra mystification; it's damn complex as is already.

-- Eugen* Leitl leitl
______________________________________________________________
ICBM: 48.07078, 11.61144 http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE
http://moleculardevices.org http://nanomachines.net





This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:43 MDT