Re: [sl4] Rolf's gambit revisited

From: Matt Mahoney (
Date: Sat Jan 10 2009 - 17:52:48 MST

--- On Sat, 1/10/09, Rolf Nelson <> wrote:

> On Sun, Jan 4, 2009 at 5:00 PM, Nick Tarleton <> wrote:
> > Besides (a) greater diversity of RAI goal systems (b) hell worlds (c)
> > ability of humans to credibly commit to do certain things if FAI is created,
> > *are there any relevant differences between FAI and RAI?*
> It's hard to know what differences would be relevant without more
> analysis. One way to start would be to begin in simple worlds with
> simple models that can be fully analyzed, and then work our way to
> more general cases.

It's hard to know without defining your terms. Unfortunately that is not possible because a description of "friendliness" is algorithmically complex (as much as AI itself). CEV tries to get around this complexity by pointing at human brains.

-- Matt Mahoney,

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT