From: Vladimir Nesov (robotact@gmail.com)
Date: Thu Mar 06 2008 - 18:54:27 MST
On Fri, Mar 7, 2008 at 4:15 AM, Matt Mahoney <matmahoney@yahoo.com> wrote:
>
> --- Bryan Bishop <kanzure@gmail.com> wrote:
>
> > On Thursday 06 March 2008, Stathis Papaioannou wrote:
> > > The recent simulation of a rat neocortical column by the IBM Blue
> > > Gene/L computer is claimed to accurately model real brain tissue.
> > > This seems almost too good to be true, but if it is, it would mean
> > > that simulation at the classical level is enough. It would also mean
> > > that simulating a human brain is only a decade or two away.
> >
> > I would agree that it's almost too good to be true, since the simulation
> > was a basic neural net simulation that you can take right out of a text
> > book instead of anything specific to rats. So that's interesting. Hm.
> > Can you link me to a ref on the accuracy of the simulation?
>
> Here is a 1 hour talk on the project.
> http://video.google.com/videoplay?docid=-2874207418572601262
>
> There was some pretty extensive and novel research leading up to the
> simulation. They found hundreds of different types of neurons within a
> cortical column and studied the behavior of each type by growing cells on
> silicon.
>
How embarrassing it would be when after 20 years of research they make
their simulation do something micey, and then find a way to simplify
neurons and connectivity with preservation of behavior to something
absurdly simple, which was known to machine learning for decades but
was never *taught* to become intelligent?
-- Vladimir Nesov robotact@gmail.com
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:02 MDT