[ExI] Human Brain Project kicks off today

[Eugen Leitl]

On Wed, Oct 09, 2013 at 12:57:26AM -0400, Rafal Smigrodzki wrote:
> On Tue, Oct 8, 2013 at 3:15 PM, Eugen Leitl <eugen at leitl.org> wrote:
> 
> > Bootstrap stages are sacrificial/discardable, just as in compiler bootstrap.
> 
> ### Ah, I am getting it now: Initial uploading/human-level AI stages
> are general purpose hardware with very bloaty software, with gigantic
> inefficiencies but then with enough specific knowledge you can build
> specialized hardware running highly optimized code. Makes sense.

Yes. See http://www.stanford.edu/group/brainsinsilicon/challenge.html
http://www.stanford.edu/group/brainsinsilicon/neurogrid.html
http://www.stanford.edu/group/brainsinsilicon/goals.html
for early work in that area. It pretty much mimicks what the
brain does, the spike shape itself encodes no information,
and the long-distance fiber structure is pretty much like
a modern supercomputer N-dimensional torus fabric, routing
packets.
 
> How many Eugens could you fit in the head of a pin (make it a sphere
> 3/16 inch diameter)?

Less than it might appear. It takes minimum nm^3 to encode one
bit in 3D solid state, so um^3 buys you 10^9, mm^3 10^18, cm^3 10^21 
bits. The actual volume is anywhere in 100 cm^3 to 1 cm^3 range.
Human brain minus water is some 300 cm^3, plus a fraction is
that for metabolism/maintenance, so the 100-1 cm^3 figure 
appears reasonable.