Re: Darwinian dynamics unlikely to apply to superintelligence

From: Tomaz Kristan (me2icq@icqmail.com)
Date: Sat Jan 03 2004 - 03:28:04 MST


On Sat, 03 Jan 2004 01:30:25 -0500, "Eliezer S.
Yudkowsky" wrote:

>
> Natural selection is a very meager optimization
> process, a poor scavenger
> of mutations. Recursive self-improvement would eat up
> all the variation
> before evolution has the chance to even look at it.
>

There is no fine line between the evolution and
intelligence. EA _is_ an intelligent algorithm.

Do some random change and wait, what is going to
happen! Let the world (environment) machine to prove,
if the change is either wrong or correct!

Later in the process some mechanisms may develop, to
early abort bad blueprints created by random.

Even later on, those changes will not be entirely
random. Some clever strategies will evolve, to promote
more and more promising mutations. But sometimes, they
will be bypassed. There is no guaranties, that the
early testing is the best possible. An unorthodox
mutation could lead to the next level eventually.

After some number of generations, one can't tell what
is this. An evolution or an intelligent design, carried
on by the evolved entities. This applies on our
biological/memetic history. It could be also applied to
some self modifying digital algorithm. Very
evolutionary at the beginning, and ever more
intelligent with passing time.

-------------------------------------------------------------
Sign up for ICQmail at http://www.icq.com/icqmail/signup.html



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:43 MDT