Re: Si definition of Friendliess

From: Eliezer S. Yudkowsky (
Date: Sat Apr 07 2001 - 11:27:51 MDT

Brian Atkins wrote:
> the reason it can't have the kind of effects we have now
> is that if you did manage to incite a riot you still couldn't kill or harm
> anyone who didn't want to be harmed. It might still be possible to have some
> real negative consequences from speech- libel perhaps? I am unsure of how
> that might be handled by a Sysop. If we assume it also protects a person's
> property, then it might also want to protect their reputation? So would it
> outright prevent libel/slander? If not, would we have to come up with some
> kind of government/social system to deal with it?

I would view "speech" as an emergent quality of volitional interactions
between citizens. Suppose that citizen A says something nasty about
citizen B to citizen C, and citizen C does something B doesn't like as a
result - this could be as simple as saying something that hurts B's
feelings, or as complex as withholding a portion of vis matter from B's
interstellar ship. The interaction between A and C is volitional, and the
interaction between C and B is volitional, so it's hard to see how the
Sysop could interfere directly.

Of course, if C asks the Sysop what ve thinks of A's comments, the Sysop
may not be able to give an answer based on A's private mindstate - but
just the fact that "A has not authorized me to discuss vis motives with
you" may be answer enough, as would "normative social perception, without
reference to private information about A's mindstate, says this is a
backstab". If the citizens don't want to trust the Sysop, then, yeah,
some kind of elaborate social system.

I expect the transhumans will be above this sort of thing, so maybe I
should have been using "he" and "she" back there. Anyway, people who
choose to remain human will never run out of ways to make themselves
unhappy. Probably the most I can do is ensure there's a choice.

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

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT