Re: purpose of AIs

Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Mon, 13 Dec 1999 14:08:05 -0800 (PST)

I think networks will be getting dumber, and more orthogonal. There will be little, if any, subvertable resources in them. You can't subvert a hardwired optical switch very much. The nodes will be getting fatter and smarter, though. That's where the action will be.

Still, you need a lot of generations up the linear log plot before you can have a god budding in your network without you noticing any significant performance degradation. I would say a Blue Gene equivalent in every game console would qualify for a supercritical broth, maybe slightly less. (Which will probably need 2 1/2 d molecular switches, if I come to think of it).

Once we're there I would tread very, very carefully: extinction alert.

Ken Clements writes:

> What we were thinking about was not desktop-sized, but rather would emerge as a
> distributed intelligence where the processing power in each router was something like a
> small cluster of neurons. As you say, the infrastructure of the net is going optical,
> which is just another example of the building of processing speed, which extends the
> upper limit on how smart this could be.