From: Ben Goertzel (ben@goertzel.org)
Date: Tue Mar 30 2004 - 03:37:01 MST
The prior will naturally "drift" due to limitations on resources leading
to inaccurate maximization of expected utility in the context of
self-modification. How it will drift depends on phenomena in the
dynamics of iteratively self-modifying systems, which none of us
understand very well yet.
I can give more details later; I have to go to a meeting now, alas ;-(
ben
> -----Original Message-----
> From: owner-sl4@sl4.org [mailto:owner-sl4@sl4.org] On Behalf
> Of Wei Dai
> Sent: Monday, March 29, 2004 11:32 PM
> To: sl4@sl4.org
> Subject: Re: escape from simulation
>
>
> On Mon, Mar 29, 2004 at 09:07:26PM -0500, Ben Goertzel wrote:
> > Since Novamente will be a fully self-modifying system (when
> complete),
> > it will be able to modify its own "prior" ;-)
>
> Ok, but how and why will it modify its own prior? I'm not
> familiar with Novamente's decision algorithm, but assuming
> that it is also based on maximization of expected utility, it
> will never *want* to change its prior, because doing so will
> always lower expected utility if EU is computed using the
> current prior rather than the new proposed prior. If it's
> computed using the new prior instead, then it will *always*
> want to change its prior, which would be an even worse outcome.
>
> If you still think Novamente will change its prior on its own
> initiative, can you please give a scenario where it might do
> that, and explain what the underlying logic of the decision could be?
>
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:46 MDT