The zdnet article has this passage in it: "The PlanetWars contest received extra attention due to a post on reddit titled “The end is near! Self-improving AI based on Genetic Programming is beating 95% of hand-coded submissions in the Google AI Challenge”. The genetic algorithm coded by a team calling itself space.invaders did better than expected but ended up in 277th place."
That really is quite an interesting development, given that the people that compete in these contests typically are pretty good at what they do, I'm really surprised that a GA derived program would do that well. Is there something about this particular challenge that sets the stage for that to happen or is there more to it?
Comments
The zdnet article has this passage in it: "The PlanetWars contest received extra attention due to a post on reddit titled “The end is near! Self-improving AI based on Genetic Programming is beating 95% of hand-coded submissions in the Google AI Challenge”. The genetic algorithm coded by a team calling itself space.invaders did better than expected but ended up in 277th place."
That really is quite an interesting development, given that the people that compete in these contests typically are pretty good at what they do, I'm really surprised that a GA derived program would do that well. Is there something about this particular challenge that sets the stage for that to happen or is there more to it?
http://ai-contest.com/forums/viewtopic.php?f=17&t=1136
I think two things:
1. The space.invaders team seemed to be good at what they were doing.
2. The "pretty good" people you refer to that also put in the required time to deliver a top bot is perhaps the top 50 (IMHO).