Re: Neural/Digital convergence: dynamic memory and bucket brigades

Date: Thu Oct 09 2003 - 20:12:52 MDT

In a message dated 10/9/2003 8:49:27 PM Eastern Standard Time, writes:

> When I went back and looked up some neuroscience papers on column and
> chain structures in the brain, I noticed that there are a lot of
> structures that look very similar to that whole family of analog buffer
> circuits, and in the places I would normally expect to find these kinds
> of buffering structures computationally. But the references to analog
> buffers I've found (and only a small number of vaguely related papers
> show up on Google), do not seem to be referring to complex multi-neuron
> structures in the way I am. Looking at some papers, it seems that they
> are looking at the same macro structures, but their thoughts are in the
> wrong place conceptually to see that model pop out.

I can see what you are talking about. We can say that a buffer is 'any method
of delaying data without losing it'; simply sending the signal down some
neural path which does nothing but return it will do the job. It's well known that
nerve signals are very slow compared to electrical impulses in a copper
fiber, and in a very large animal (a sauropod dinosaur, for instance) it would take
a good chunk of a second for a pain signal from an injury to the tip of the
tail to reach the brain 60 feet away. This is probably why these animals had
large nerve bundles farther down in the hip area. Even in humans reflexes are
much faster for hand controls than foot pedals, which is why one car company is
developing a vehicle with all controls on the wheel.
So column structures in the brain can make a good buffer. I presume the
delays are on the order of hundredths of a second, given the distances in the
brain. In developmental terms, the structures probably sort themselves out in the
prenatal environment, and probably have existed in some form for about as long
as animals have found them useful; I wouldn't be surprised to hear that some
worms probably had buffers half a billion years ago, as soon as they had a few
hundred neurons to play with.
Incidentally, I have wondered whether the human brain may not depend on sound
for a sort of 'basic programming'.
In utero, no other sensory component is really operative yet (AFAIK), but the
constant rhythm of the mother's heart, among other sounds, would seem to
comprise almost the whole of the sensorium and, in a sense, almost the only
reality the fetal brain has to work with.
It seemed to me that one possible approach to an AI might be to set up a
neural network and develop it in a way fashioned after what we know about human
brains, beginning with only aural input and a sort of sensorimotor homunculus,
as opposed to simplified visual inputs. Indeed, I spent the better part of an
evening toying with the idea of using a robotic baby as the interface, with
fully motorized limbs, visual inputs in the eyes, tactile sensors in the skin,
and so forth, and the computer itself located in this body or elsewhere with
wireless links; the programmers would deal with the AI by literally raising the
baby as best they could.
Tom Buckner

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