[extropy-chat] Mouse brain simulated on computer

[Eugen Leitl]

On Sat, Apr 28, 2007 at 06:44:31AM +0000, Andrew Burrows wrote:
> News article is linked here: 
> http://news.bbc.co.uk/1/hi/technology/6600965.stm
> and after some searching, here is the official site with papers: 
> http://brain.cs.unr.edu/index.php
> 
> Don't get too excited from the sensationalist title. It's not quite a whole 
> brain simulation, just 8,000 neurons. However the researchers are trying to 

As usual, the reporters didn't bother to check their numbers for
plausibility. It was 8 MNeurons, on a 4 kNode Blue Gene/L.
The Blue Gene topping Top 500 is a 64 kNode system, which,
assuming the simulation would scale, would give you 8 mice,
or one mouse run at about realtime.

Of course, that simulation is rather removed from biologic
realism, but it is a nice benchmark for spiking codes for
AI applications.

The Blue Gene family is designed to scale to some 2 MNodes,
and future models will be based on the Cell CPU.

> be faithful to the real world interactions of neurons.
> 
> And presumably with this model it can help deduce simpler neuron 
> representations, further increasing efficiency. I won't expect any practical 
> applications to come from this for many years at least due to the 
> computational expense of the model.

A 4 kNode Blue Gene/L is a low end model, as far as scientific
supercomputers go. It's just two racks of hardware. Right now
you could buy a few PS3, install Linux + MPI on them, and hook 
them up to a large GBit switch. You'd get a 48-node/12 GByte 
Cell cluster for the low, low price of 25 k$. And it would
scale as far as your electricity bill could carry.

-- 
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