Re: Revising a Friendly AI

From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Mon Dec 11 2000 - 14:05:59 MST


Samantha Atkins wrote:
>
> Oh yeah. I've had both actual and hypothetical opportunities to play
> junior goddess or at least local guru in various contexts. It is not
> too difficult to see how to majorly engineer some aspects of local (or
> even more global) human culture in some circumstances. Frankly the
> responsibility without checks and given my own self-knowledge scares the
> bejeezus out of me. Of course, sitting back from the fray and watching
> what other less inhibited players do or usually worse, what just evolves
> naturally, isn't much fun either.

Do you think that, if you had the ability to modify your own mindware, you
could transform yourself into something you could trust? If so, then
there's the *seed* of something trustworthy in the morality you already
have, and it should be theoretically possible for you to create a
trustworthy AI - maybe just by leaving some of the yuckier parts out.

-- -- -- -- --
Eliezer S. Yudkowsky http://intelligence.org/
Research Fellow, Singularity Institute for Artificial Intelligence



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:35 MDT