Two Carnegie Mellon University students have programmed the ultimate killing machine. If you face off against their AI in Doom’s deathmatch, that is.
As originally reported by NextBigFuture, this new bot (affectionately named Arnold) uses a host of complex mechanisms to play Doom better than any human possibly can. Created by Guillaume Lample and Devendra Chaplot, the bot uses the same networking that allowed Google’s DeepMind to learn how to play Atari games.
But the jump from 2-D to 3-D provided unique challenges. Traditional AI bots rely on information that players can’t possibly have such as map data and other in game variables. Lample and Chaplot aimed to teach Arnold how to play while only reacting to what was visible on screen. It’s a bit complicated but the bot has a functional short term, long term memory to help it keep track of what it encounters.
Still, it takes a long time for AI to learn how to play games. Bots that iteratively learn through neuro-evolution can take days to make significant progress. A bot learning to play a single level of Super Mario World took 24 hours. Arnold made little progress after an initial 50 hours of play. His creators had to use an API to access Doom’s engine to help speed along the learning process.
Arnold took second place at the Visual Doom AI Competition, even though it did manage to have the highest kill to death ratio. A bot named F1 managed to get more frags.
If you’re interested in the particulars, you can also read Lample and Chaplot’s paper on the AI. Me? I’ll be cowering in a corner, hoping that fragging enemies in Doom will be enough to sate Arnold and his deadly skills.