> So give that you don't know how much the human brain does crunch
> and how effectively it work, can you quantify "far more effective"?
> And give the reationale behind that estimate?

I can't give the full rationale in a brief email and don't have time
to write an essay on it right now.

But in brief, *if* the Novamente architecture as currently formulated
is workable then the hardware req's should be around 500-1000 current
PC's networked together.  This is based on the idea that we will need
to automatically learn "combinator tree" programs in Novamente's
internal programming language with roughly 500 nodes in them, via
probabilistic evolutionary learning where the fitness evaluation uses
Novamente PTL inference on a single machine.  And it's based on some
loose curve-fitting to guess how much hardware power is needed to
learn combinator trees of this size.

Of course there is a lot of reasoning here which I am not presenting
in this email -- but anyway, yeah, I do have an estimate which is
based on detailed reasoning...

ben

-------
To unsubscribe, change your address, or temporarily deactivate your 
subscription, 
please go to http://v2.listbox.com/member/[EMAIL PROTECTED]

Reply via email to