[extropy-chat] proto-extropian religions

Eliezer Yudkowsky sentience at pobox.com
Wed Oct 13 05:48:25 UTC 2004


The Avantguardian wrote:
> Creationism and evolution
> are both ridiculously low resolution maps of the
> ultimate truth.

Suppose I flip, one million times, a coin biased toward 90% heads.

To say that the coin comes up heads 9 times out of 10 is a ridiculously 
low-resolution map compared to the actual sequence.  In Bayesian terms, the 
final probability our hypothesis assigns to the actual sequence will 
probably be around ((0.9)^900,000)*((0.1)^100,000), an infinitesimally tiny 
number.  This tiny number is a terribly long way from 1.0, the unattainable 
Truth, the score of correctly stating the exact sequence in advance as your 
sole prediction.

Nonetheless, the hypothesis that the coin comes up heads 1 time out of 10, 
also a ridiculously low-resolution map, scores far worse than the 
hypothesis that the coin comes up heads 9 times out of 10.  The 1/10 
hypothesis will score far worse than the hypothesis of maximum entropy 
(that we are entirely uncertain, 50/50, whether the coin comes up head or 
tails on any given flip).

And so creationism and evolution are both ridiculously low-resolution maps, 
but evolution scores vastly better than the hypothesis of maximum entropy 
while creationism scores vastly worse.  In an uncertain universe, this is 
the closest we come to defining a "correct hypothesis" and "incorrect 
hypothesis".

-- 
Eliezer S. Yudkowsky                          http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence



More information about the extropy-chat mailing list