From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Thu Sep 10 1998 - 19:02:49 MDT
Joe Jenkins wrote:
>
> From: Eliezer S. Yudkowsky <sentience@pobox.com>
>
> > Also, the first-stage neurological transhumans will just create AIs,
> > since it's easier to design a new mind than untangle an evolved one.
>
> By "design a new mind", do you mean evolve a new mind using genetic
> algorithms?
No, I mean that someone has to sit down and think until he understands what
the brain does and how, and then deliberately program it into a computer.
> If so, why would that be easier to untangle?
It wouldn't, which is why I specifically say in "Coding a Transhuman AI" that
neural networks should be used as a last resort.
> Also, why
> must the naturally evolved mind be completely untangled rather than
> just the interfaces?
So that basic improvements can be made, not just surface links to pocket calculators.
> And why can't the naturally evolved mind be
> uploaded and evolved further using genetic algorithms - a starting
> point for the GA?
Uploading requires nanotechnology on such a high level that someone has long
since reduced the Earth to a boiling puddle of goo.
--
sentience@pobox.com Eliezer S. Yudkowsky
http://pobox.com/~sentience/AI_design.temp.html
http://pobox.com/~sentience/sing_analysis.html
Disclaimer: Unless otherwise specified, I'm not telling you
everything I think I know.
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:49:34 MST