On Tue, Nov 19, 2002 at 08:13:52AM -0600, The Fool wrote:
> Hrm.  If you include all the shorter pathways it's
> 
> the sum of N! + (N-1)!...2!
> 
> 3! + 2! = 8

None of which is relevant to the question of processing power of the
brain.

If you approximate the brain as a bunch of synapses (5E14, say)
connected between a bunch of neurons (1E11), then each synapse needs
about 3 bytes of storage space to store the address of two neurons (with
1 bit to indicate whether the synapse is on or off). That comes to
1.5E15 bytes. So, if you do 3 operations per neuron as you go through
one update cycle of your state machine, then you need about 1E15 ops per
second, so IBM's machine falls a little short, but it is close, probably
withint the range of error of this approximation. Accessing the data
structure as memory elements would slow this down a lot, though, but I
bet some clever computer scientist could come up with a near-optimal
clustering of the neurons so that a lot of the data could stay in the
registers.


-- 
"Erik Reuter" <[EMAIL PROTECTED]>       http://www.erikreuter.net/
_______________________________________________
http://www.mccmedia.com/mailman/listinfo/brin-l

Reply via email to