X-Message-Number: 26054
Date: Mon, 18 Apr 2005 20:45:52 -0400
From: Daniel Crevier <>
Subject: To Tom Donaldson on brain and weather simulation
References: <>

To Tom Donaldson, on brain & weather simulation

You can't blame the inaccuracy of weather forecasts on the fact that they
are performed on digital computers. The real reason is the well known
'butterfly effect': the extreme sensitivity of air mass dynamics to small
disturbances. These disturbances can be small inaccuracies in initial
conditions, or small perturbations like, yes, butterflies (although
airliners and power plant exhausts are probably more to blame). Increasing
the accuracy of the computations themselves would make no difference: it is
our own knowledge of the details of the system that is at fault.

This means that you could replace present weather computers, which
substitute, as you say, real simultaneity with 'make-believe simultaneity',
by very many analog computers, one per cell in the simulation. These analog
computers would interact with perfect simultaneity, exactly like the neurons
in a brain. Yet, if you use the same intial conditions and don't include
every living butterfly or airliner, the results would be just the same as
those obtained with a digital computer.

Weather simulation is thus a legitimate example of digital computers
simulating a very complex distributed system faster than real time, and just
as accurately as would an analog and perfectly parallel simulation.
Therefore it is not clear that modelling a brain with hard-wired neurons
would be any more accurate or economical than with digital computers.

Daniel Crevier

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=26054