Posted by
Marcus G. Daniels on
Apr 05, 2007; 3:37am
URL: http://friam.383.s1.nabble.com/One-of-my-projects-tp523626p523651.html
Phil Henshaw wrote:
> how do you represent the systems of nature that are out of control and
> making up altogether new rules???
At some point that kind effort is less of an empirical science and more
of a mathematical investigation into worlds as they could be. That's
not to say it is bad, it's just a different goal.
One way to proceed with that kind of investigation is with genetic
programming. Create an imaginary world that has certain forces acting
on the things in it, and then evolve computer programs that can survive
in that imaginary world. After the agents survive very well, take
apart those computer programs to try figure out how they work, or study
how different computer programs interact in that world and possibly even
change it. Classic example:
http://www.archive.org/details/sims_evolved_virtual_creatures_1994With an avatar/gaming world, it's not hard to imagine automated agents
learning how to fight or cooperate with human players. Then one could
probe those agents to watch how they make decisions. To be more
systematic and learn about learning one could have timestamps on each
node/branch to compare the recent innovations from enduring logic.