Re: DiscoveryCh - AI

From: Eliezer S. Yudkowsky (
Date: Wed Jan 24 2001 - 16:02:40 MST wrote:
> I do not believe in super-docile AIs. The AI we can build
> and control is next to useless (good for sorting trash,
> delivering goods, keeping the house tidy and driving a car),
> and a useful AI is uncontrollable.

Just a brief note: "Uncontrollable" is, semantically, orthogonal to
"friendly". Martin Luther King and Gandhi were neither docile nor
controllable, but they were altruistic. You can, of course, argue that a
smart AI is necessarily unFriendly as well as uncontrollable, but we
should all be careful not to conflate the concepts and add confused
terminology to an already-chaotic issue.

We now return you to your previous argument.

-- -- -- -- --
Eliezer S. Yudkowsky
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:56:24 MDT