AI Challenge: Ants

lavalamp

Aichallenge.org has started their third AI contest this year: Ants.

The AI Challenge is all about creating artificial intelligence, whether you are a beginning programmer or an expert. ... [Y]ou will create a computer program (in any language) that controls a colony of ants which fight against other colonies for domination. ... The current phase of the contest will end December 18th at 11:59pm EST. At that time submissions will be closed. Shortly thereafter the final tournament will be started. ... Upon completion the contest winner will be announced and all results will be publically available.

Ants is a multi-player strategy game set on a plot of dirt with water for obstacles and food that randomly drops. Each player has one or more hills where ants will spawn. The objective is for players to seek and destroy the most enemy ant hills while defending their own hills. Players must also gather food to spawn more ants, however, if all of a player's hills are destroyed they can't spawn any more ants.

I mentioned this in the open thread, and there was a discussion about possibly making one or more "official" LessWrong teams. D_Alex has offered a motivational prize. If this interests you, please discuss in the comments!

Aichallenge.org has started their third AI contest this year: Ants.

The AI Challenge is all about creating artificial intelligence, whether you are a beginning programmer or an expert. ... [Y]ou will create a computer program (in any language) that controls a colony of ants which fight against other colonies for domination. ... The current phase of the contest will end December 18th at 11:59pm EST. At that time submissions will be closed. Shortly thereafter the final tournament will be started. ... Upon completion the contest winner will be announced and all results will be publically available.

Ants is a multi-player strategy game set on a plot of dirt with water for obstacles and food that randomly drops. Each player has one or more hills where ants will spawn. The objective is for players to seek and destroy the most enemy ant hills while defending their own hills. Players must also gather food to spawn more ants, however, if all of a player's hills are destroyed they can't spawn any more ants.

I find it quite hard to believe you couldn't do even better if you were a single mind perceiving what the ants did and controlling them (which is how you are set up in this game). A single mind can, worst case, simulate the rules each ant follows, so it can never be worse than the social behavior is. But the ants individually can't simulate a single large mind (for one thing, they wouldn't have all the information it would have).

It'd be like writing a chess engine by writing a different AI for each piece. Splitting up the AI gives you more problems, not less.

(That's not to say that you couldn't evolve a good set of local rules to follow in this game, of course!)

I find it quite hard to believe you couldn't do even better if you were a single mind perceiving what the ants did and controlling them

Sorry for the additional response to the same post, but I feel this bears special notice, as I just now realized that we might be talking "past" one another.

IF the purpose of the endeavor isn't just to "do better", but rather to learn about how intelligence and cognition operate, then it seems to me that examining a real-world manifestation of intelligence (and even cognition in the form of observing... (read more)

-1Logos0115y

You heard about the Berkeley Overmind? A single mind has a limited amount of capacity to focus on events simultaneously. Emergent intelligence does not. [...] Given the total neural processing power available to the ants, I'd dare say that their capacity to solve problems is far greater than you're giving them credit for. Also, there's a non-trivial chance that this is already how individual minds operate -- I'm speaking of the Society of Mind hypothesis.

22

AI Challenge: Ants

22

22

22

AI Challenge: Ants

22

22