The race between Nanotechnology and AI

Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Sun, 25 Jul 1999 22:07:04 -0700 (PDT)

John Clark writes:

> It seems a little early to me but let's assume you're right and
> Nanotechnology becomes possible in 2015 , uploading would be pretty
> straightforward after that, I don't think you'd have long to wait.

I disagree. You can breed superhuman AI from scratch much easier than to make an upload. Building optimal, or near-optimal nanocomputers is very easy, and an evolved AI could certainly utilize the optimal hardware optimally.

Uploading requires either low-artefact brain vitrification and then destructive scan or gradual in vivo uploading, with 10^6..10^9 nanomachines operating within your cranium. Doing this will be hard for us, but trivial to an AI. However, the AI probably has absolutely no interest in doing our housework for us.

> We've already seen a smaller gap between supercomputers only
> huge government agencies could afford and machines of equal power
> that third graders get on their birthday. You haven't seen anything yet,

The gap is only temporal: the supercomputers might use the same processors, but I don't know very many 3rd-graders who own ~6 kCPU machines, along with the according infrastructure.

> In the Nanotechnology age almost any endeavor our puny non Transhuman
> brains can now consider of could be put into one of two categories,
> dirt cheap or impossible.