[ExI] Kelly's future

Eugen Leitl eugen at leitl.org
Fri May 27 09:58:44 UTC 2011


On Thu, May 26, 2011 at 03:49:59PM -0700, Damien Sullivan wrote:

> I'd guess Samantha's just talking about modeling the brain as an
> abstract neural network, with each neuron being modeled simply as its
> synaptic weights, never mind organelles or molecular activity.  I figure
> it would take 100 million desktop-PC equivalents to model the human

Desktops are a poor way to turn Joules into Ops. If it has to
be a generic computer you'd probably cramming as many aircooled
ARM SoCs on a mesh into a rack as possible. 

Which would be still a far cry from a dedicated ASIC, nevermind
a hybrid box.

> brain like that, which at 20 watts per PC would be 2 GW.  And I've read
> a modern workstation can be more like 400 W.  Actual modeling the
> insides of the neurons, I can't even estimate that.
> 
> Meanwhile, the actual human brain operates on like 15 W.

At 50 EUR/MWh a MW will only set you back by 1200 EUR/day.
So more than 10 MW for a human equivalent will be probably
exceptional, unless power gets a lot more cheaper.

-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE



More information about the extropy-chat mailing list