From: Tim Freeman (firstname.lastname@example.org)
Date: Wed Aug 22 2007 - 21:24:26 MDT
From: "rolf nelson" <email@example.com>
>If my RAI* moves ahead and attempts to destroy mankind, I will punish
>it by pulling the plug such that RAI* will never calculate C. ...
>Therefore, it *may* be deterred by my threat.
>(This probably depends on what philosophical beliefs it implicitly
>holds about probability.)
Right. One reasonable philosophical belief is the universal prior.
(See http://en.wikipedia.org/wiki/Inductive_inference, except most
people here already seem to know about it.)
If the RAI assumes the universal prior, and it takes at least 1000
bits to transform an explanation of the world that says "these are the
laws of physics" into one that says "these are the laws of physics
because this simulator run by people with these motives computes it",
then the RIA will estimate the probablity of the simulation scenario
to be one in 2 ** 1000.
I don't think you'll get deterrence to happen in this scenario, unless
the cost to the AI of preserving mankind is essentially zero.
Is there a reasonable a-priori distribution that gives nonnegligible
probabilities to simulation scenarios?
-- Tim Freeman http://www.fungible.com firstname.lastname@example.org
This archive was generated by hypermail 2.1.5 : Fri May 24 2013 - 04:01:06 MDT