It seems to be a popular thing lately for programmers to develop AIs capable of learning to play games.
[QUOTE=opaali;47216710][media]http://www.youtube.com/watch?v=YGJHR9Ovszs[/media] ?[/QUOTE]
Same principle but works best on side scrollers best and requires training data unlike OPs
[QUOTE=Ylsid;47216987]Same principle but works best on side scrollers best and requires training data unlike OPs[/QUOTE]
But is OP's AI a [URL="http://youtu.be/Q-WgQcnessA?t=13m20s"]pinball wizard[/URL]?
I didn't think so.
[sp]Also Playfun can subtly manipulate the RNG in a game just by playing a game in a specific way it figures out[/sp]
I assume the AI accesses the memory of the game so it knows the game-states and score increments? It'd be even cooler if it has some sort of visual learning base, where it merely observes the screen-buffer and figures things out from there so it could technically be applied to anything.
[QUOTE=SweetSwifter;47217584]I assume the AI accesses the memory of the game so it knows the game-states and score increments? It'd be even cooler if it has some sort of visual learning base, where it merely observes the screen-buffer and figures things out from there so it could technically be applied to anything.[/QUOTE]
That is extremely difficult, it would require tech far more advanced than that used in driver-less cars.
Just don't teach it to play Missile Defense
super meat boy when?
[QUOTE=Minelayer;47220837]super meat boy when?[/QUOTE]
they're trying to develop it, not make it want to kill itself
[QUOTE=Octopod;47222900]they're trying to develop it, not make it want to kill itself[/QUOTE]
He said Super Meat Boy, not I Wanna Be The Guy
Machine learning is neat and all, but far from actually resembling intelligence. A computer builds a model; humans build an understanding.
For example, when a computer learns to play the first level of Super Mario World, it might be able to find a great strategy for finishing it with a high score after hundreds of training cycles. But if you then put it in the second level, it has to start all over again.
My dad was telling me about this recently, glad there is a video showing it, that is pretty impressive
It only took like 600 trial and errors to go pro
"A strange game. The only winning move is not to play. How about a nice game of chess?"
i wonder how good it would do in a game of DEFCON
Sorry, you need to Log In to post a reply to this thread.
