Re: artificial evolution of AI

From: Martin Striz (metastriz@yahoo.com)
Date: Tue Feb 01 2005 - 10:55:02 MST


--- Thomas Buckner <tcbevolver@yahoo.com> wrote:

> I'm a longtime subscriber (I buttonholed Marvin
> Minsky with a muscle-wire suit proposal at the
> June 2001 Discover best-inventions-of-the-year
> awards in NYC). This is a great article, about
> scientists running a program which is 'not a
> simulation, but an actual example' of evolution,
> and which has demonstrated that many of the
> things creationists consider improbable are
> actually quite probable. For example, even
> primitive eyes are useful, and complex ones
> evolve easily. The link Martin offers is for that
> software. Been meaning to go check it out...

Indeed, the article is far more interesting as a vindication of evolution, but
I figured, if you set the selection pressure for "computing power" or something
(and the agents in article are already doing simple number crunching), then you
could, in principle, evolve AI. All you'd need is ton of processing power and
a lot of time.

Also, I loved how it demonstrated that irreducible complexity could evolve (a
favorite for Intelligent Design proponents), for example, the program with 19
instructions, which doesn't work if missing any one of them.

Martin

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:50 MDT