Re: poly: Recombinant Growth

Robin Hanson
hanson@econ.berkeley.edu http://hanson.berkeley.edu/

> I like this paper. It suggests that new ideas are just
> combinations of old ideas, so the supply of new ideas is
> unlimited, and the fundamental limit to economic growth rates
> is the cost to examine each new idea.

I recall reading an idea several years ago, possibly by Hofstadter,
for mechanizing creativity. He suggested dividing it into two parts,
one which randomly created many possible ideas (or artworks, etc.) and
the other which evaluated them. You could also have these interact so
that the random creation could be guided to focus within certain areas
to find locally optimal designs.

It was not clear though that the random creation process would produce
a high enough concentration of good ideas that this plan will work.
Genetic algorithm schemes usually have to be carefully designed so
that the evolution rules will produce something which is effective,
something which preserves some of the properties of the parent(s).
Otherwise you just keep jumping around at random within the solution
space and never get anything good.

Hal
Received on Fri Aug 7 21:34:43 1998