Re: supergoal stability

From: Wei Dai (weidai@eskimo.com)
Date: Sun May 05 2002 - 00:11:11 MDT


On Fri, May 03, 2002 at 05:14:35PM -0600, Ben Goertzel wrote:
> The selfishness of the individual creating the AI is not necessarily
> correlated with the selfishness of the AI. A selfish person might well
> create an AI that valued humans over itself -- while putting in a special
> clause to make it value its creator most of all!

I agree with that. My question was, if some selfish person was to try to
build an AI that favors the creator, can the supergoal remain stable as
the AI modified itself? I assume that no one would be stupid enough to try
to build a purely selfish AI. (That may be the unintentional outcome but
it seems unlikely that anyone would have it as the goal.)

What kind of AI are you trying to build? A uniformly altruistic one, like
Eliezer's Friendly AI, or a non-uniformly altruistic one that favors some
people over others?



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:38 MDT