Videos

Computer program that learns to play classic NES games



suckerpinch

This is an explanation and demo of software I wrote that learns how to play a Nintendo Entertainment System game and then automatically plays it. This is real.

Research paper published in SIGBOVIK 2013: “The first level of Super Mario Bros. is easy with lexicographic ordering a and time travel …after that it gets a little tricky.”
http://tom7.org/mario/mario.pdf

Follow-up video with more games: http://www.youtube.com/watch?v=YGJHR9Ovszs
And episode 3: http://www.youtube.com/watch?v=Q-WgQcnessA

For more info and source code, see: http://tom7.org/mario/

Source

Similar Posts

45 thoughts on “Computer program that learns to play classic NES games
  1. Should we not be sort-of-scared that, when taught the importance of "not losing" – the 'A.I' chooses to forever and eternally "win" by causing a stalemate?

    Would this not ensue against our favour, come time for Real Life™ Warfare with future A.I.?

    In a way, we should be very scared, perhaps… as one day, there will be a force we will be up against, that will choose the highest negative outcome, merely so that it does not "lose" the conflict…

    [Note that the above is fun, sci-fi babble, not taking into account actual programming/coding limitations and functionality or artificial intelligence capabilities in a real, live scenario]

  2. I'm concerned about people who think that Playfun pausing Tetris is creepy. I mean, it's programmed to avoid losing, right? And it knows how to pause the game, and that game thingies not move when paused. So when Playfun drove itself into a situation where it was about to lose, it went for the only option that would allow it to not lose (as long as the game is paused). It's logical. And kind of adorable.

  3. "Hudson" is two Japanese dudes who started an electronics shop and liked trains so named their shop after Hudson locomotives, their favorite brand. The shop turned into a software/game developer and eventually sold to Konami.

  4. Let's all comment on the last second of the video (except credits you skipped) and ignore the rest.
    "My attention time span is 1 second long. Remember that part where it paused? SkyNet, amiright?"

  5. I suggest you study the Chessmaster AI. You simply need to think ahead. So that the CPU knows, if it goes here, it loses, falls in a pit, or gets hit by an enemy. My point is, if you can teach a computer to play Chess, you should be able to teach a computer to play any other game.

Comments are closed.

WP2Social Auto Publish Powered By : XYZScripts.com