> From: [EMAIL PROTECTED]
> 
In a message dated 3/4/2002 12:49:10 PM Eastern Standard Time,
[EMAIL PROTECTED] 
writes:


> Processing at the same bitrate and simulate the human mind are
different
> things.  What they have to do to simulate a human mind is to simulate
> it's, connectivity, its network.  Each Neuron has up to something like
> thousands of different connections.  In whole the human brain has more
> possible connections then there are sub-atomic particles in the
universe.
> 
> The neuron can respond in only a binary manner (fire not fire =
depolarize  
> not depolarize) but the number of influences on the state of the neuron
are 
> so great that predicting when a neuron would fire and altering the
firing 
> potential would be almost impossible. the feedback loops and
biochemical 
> interactions are obscenely complicated. About 3 months ago I heard one
of 
> the Noble winners from Rockefeller U give an "abbreviated"  review of
the 
> neuronal function. Very elegent powerpoint presentation. Within 10
minutes 
> my head was spinning and he had just talked about the complexity of the

> excitory pathways only mentioning briefly how the inhibitory pathways 
> worked and how each pathway effected each other at mutliple sites along
the 
> enzyme chain.

With enough studying they will eventually be able to replicate a single
neuron in software.  Once that has been accomplished (and a great
accomplishment it will be) they will start connecting more and more of
these simulated neurons together, not neccesarrilly in the same
organization as a human mind.  (Undoubtedly they will use some sort of
genetic programming techniques to develop the mind of the simulated AI). 
It may be that it takes a significant amount of processor speed / memory
to accurately simulate a single neuron (like say a top of the line Xeon
computer).  This would require connecting a lot of computer hardware
together.

Reply via email to