From: Mitchell Porter (firstname.lastname@example.org)
Date: Sun Jan 28 2001 - 19:53:32 MST
It seems to me that the basic principle is
Whatever goals you want your super-AI to have,
make them subgoals of a Friendliness supergoal.
And that the important observation which backs
this up is that survival and intelligence are
natural subgoals of Friendliness, because
(i) if it's dead, the AI can't be actively Friendly
(ii) if it's dumb, the AI can't be effectively
So a Friendly AI is not *obviously* disadvantaged
with respect to 'survivalist' or 'scientist' AIs.
At any rate, this is the thirty-second summary
I will use if I ever have to explain what
Friendliness amounts to as a design principle.
Unless someone can give me a better description.
Do You Yahoo!?
Yahoo! Auctions - Buy the things you want at great prices.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:35 MDT