Re: [agi] A difficulty with AI reflectivity

From: Christian Szegedy (szegedy@or.uni-bonn.de)
Date: Thu Oct 21 2004 - 05:49:27 MDT


Eliezer Yudkowsky wrote:

> Schmidhuber spoke of proving only that the code modification had
> greater expected utility. Even if you proved the new system would do
> exactly what the old system did, only faster, it would not follow that
> the new system was an *improvement*.

It seems to me that Schmidhubers algorithm is an improved version of
the Hutters universal search algorithm. A very intuitive, but good
explanation can be found on this page:

http://fortnow.com/lance/complog/archive/2003_05_04_archive.html

(Look for universal search.)



This archive was generated by hypermail 2.1.5 : Tue Feb 21 2006 - 04:22:47 MST