X-Message-Number: 14926
Date: Thu, 16 Nov 2000 00:48:28 +0100
From: Henri Kluytmans <>
Subject: computers and brains

-->

Thomas Donaldson argues that realtime simulation of the inter-
connections of biological neurons is a serious problem for 
Turing machines. He seems to suggest that the problem arises 
because all possible connections between all neurons 
should be taken into account. And this number increases 
enormously (like (n-1)! ). However it appears that the major 
part of all interconnections between neurons in the brain are 
relatively local. This quite diminishes the connective complexity 
required. 

I also must mention again that it is fairly unlikely that the 
hardware architecture for uploading a human mind will be like 
a Turing machine.

Furthermore he argues that current digital neural nets are 
only partial models. I agree with that. But does he dare to 
claim that we will not succeed in creating a sufficient model 
within a hundred years, or a thousand years, or ever ?

The question should be : "Will the dynamics of the interconnections 
be implemented in hardware or software ?" 

There is always the option to represent every change in the connections 
by a real change in the hardware architecture, using this method there 
shouldn't be any more problems than in biological hardware.

But because the major part of interconnections in our brain is 
relatively local it's likely that there are no insurmountable 
problems with implementing the connection changes in software, 
and not in hardware, because the hardware option will be less 
energy efficient (and also slower).

Grtz,
>Hkl

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=14926