AI will frag each other with rocket launchers in 'Doom'
Open competition is the next step in visual learning for our future overlords.
An AI learning to walk through a Doom-inspired maze by sight is one thing, but how can it handle live multiplayer mayhem? That's what the "Visual Doom AI" competition this September hopes to discover. The first set of matches are limited to a dozen 10-minute rounds on a known map, with only one weapon: the rocket launcher. The AI "controllers" can pick up health packs and ammo, as well, and the winner will be picked by highest kill count.
The other round mixes it up a bit by allowing multiple weapons and items in full deathmatch on a trio of unknown maps. This is incredibly important for machine learning because rather than typical bots in a game, the controllers here don't have access to the underlying code or map layout -- everything is picked up by visual learning.
Submission deadline for the warmup round is May 31st, and the final competition will take place at this September's Computational Intelligence and Games symposium in Greece. The real question here, though, is how long it'll take before DeepMind starts camping with the BFG.
