[ml] AI competition: play Mario Brothers

[Paul Oppenheim]

So,

http://ice-gic.ieee-cesoc.org/competitions.html

and

http://julian.togelius.com/mariocompetition2009/

describe that you can do anything, learning or no.

this is a lot like the final project I did for

http://www.lsa.umich.edu/saa/transfer/courseguide/archive/wint03tscg/252.html#anchor252498001

which I mentioned,

http://pauloppenheim.com/portfolio/bogwars.html

writing a GA to evolve a neural net sounds great and all, but the team 
that kicked ass wrote a state machine, and had an easy time of it, IIRC.

So, I'm interested, but this certainly would fall in the realm of 
spreading myself thin.

+ paul

Josh Myer wrote:
> For the people who have done some AI/ML work before: there's a
> competition on to build systems to play a version of Super Mario
> Brothers.
> 
> Here's the fluffier Popular Science piece:
> 
> http://www.popsci.com/scitech/article/2009-08/humans-challenge-ai-game-mario-bros
> 
> And a less fluffy New Scientist piece, with actual links:
> 
> http://www.newscientist.com/article/dn17560-race-is-on-to-evolve-the-ultimate-mario.html?DCMP=OTC-rss&nsref=online-news
> 
> 
> I'd be kind of interested to hack on this, but there's no way I could
> help get something serious done by either deadline.  Either way, it's
> a really fun thing to think about, and might be fun to hack at.
> 
> Is anybody else excited about this, and have concrete ideas of how to
> start tackling the problem?  (Sorry, I'd rather not try to help
> newbies along too much in this process, as it's, err, feckin' hard =)