Re: Human-level CPU power crossover date

From: Durant Schoon (
Date: Thu Apr 12 2001 - 17:53:25 MDT

> From: James Rogers <>
> In other words, I believe that AI on silicon is quite feasible without
> requiring huge jumps in technology. However, the software implementation
> should be appropriate for the hardware; putting a human brain on silicon
> would seem to be grossly inefficient and hardly required.

Hardly required for AI, but I want AI *and* uploading :) Who knows maybe
some of the wicked fast DSP chips you mentioned would be really good at
the processing the confocal(*) mapping in the brain that Eliezer mentioned.

But yes, I'm in your camp that AI on other types of architectures should
be possible (give Danny Hillis enough tinker-toys and maybe he'll do it :)
I don't mind waiting a few extra years for an FAI running on a Beowulf
cluster to solve all the technical upload problems. I just wonder how much
it's going to cost me...

(*) A mathematician friend once described his reasearch to me on confocal
mapping and how experiments indicated that the brain's mappings are indeed
confocal (you can google "confocal maps" for the mathematics, it's been a
while since I've heard all the details).

Durant Schoon

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT