X-Message-Number: 26896
From: 
Date: Sun, 28 Aug 2005 08:13:27 EDT
Subject: Re: CryoNet #26882 To Donaldson

T. Donaldson said:

Any  attempt to imitate a whole working brain must deal with the
speed of  processing of which it is or isn't capable. We can certainly
produce very  slow simulations of networks of neurons; and these
would certainly be  interesting as objects to study just how our
brain's neurons work. However  they would be useless as devices 
able to act like real brains in the real  world, precisely because
of the problem of TIMING.

In short, your  comments forget or (deliberately?) omit the problem
of TIMING, which for  any true imitation of a brain must match.



OK, so, you accept now that a sequential processor  is equivalent to a  

multiprocessor in the range of problem that it can work out. The difference  
being 
that the multiprocessor is faster. So we agree on that with yet one  precision 
: A (good) multiprocessor is faster than a single one * at equal  

technology*. Today single processors are faster than multiprocessors 25 years  
ago for 
example. About the brain, the minimum clock time is on the order of the  

millisecond, the comming pseudomorphic HEMT technology (used for TVsats  
reception) 
will bring processor clocking in the then of GHz range, ten millions  times 
faster that a neuron. That may reduce the requested parallel  processing...
 
Yvan Bozzonetti.
 


 Content-Type: text/html; charset="US-ASCII"

[ AUTOMATICALLY SKIPPING HTML ENCODING! ] 

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=26896