Tic tac toe, checkers, chess, go, poker. Artificial intelligence rolled over each of these games like a relentless tide. Now Google’s DeepMind is taking on the multiplayer space-war videogame StarCraft II. No one expects the robot to win anytime soon. But when it does, it will be a far greater achievement than DeepMind’s conquest of Go—and not just because StarCraft is a professional e-sport watched by fans for millions of hours each month.
DeepMind and Blizzard Entertainment, the company behind StarCraft, just released the tools to let AI researchers create bots capable of competing in a galactic war against humans. The bots will see and do all all the things human players can do, and nothing more. They will not enjoy an unfair advantage.
DeepMind and Blizzard also are opening a cache of data from 65,000 past StarCraft II games that will likely be vital to the development of these bots, and say the trove will grow by around half a million games each month. DeepMind applied machine-learning techniques to Go matchups to develop its champion-beating Go bot, AlphaGo. A new DeepMind paper includes early results from feeding StarCraft data to its learning software, and shows it is a long way from mastering the game. And Google is not the only big company getting more serious about StarCraft. Late Monday, Facebook released its own collection of data from 65,000 human-on-human games of the original StarCraft to help bot builders.
Such efforts could produce more than just fun. Google says it used machine learning from DeepMind to slash cooling bills in company datacenters. Mastering StarCraft could see software take on more complex and lucrative jobs. “From a scientific point of view, the properties of StarCraft are very much like the properties of real life,” says David Churchill, a professor at Memorial University of Newfoundland who advised DeepMind on its StarCraft tools, and organizer of a leading StarCraft bot competition. “We’re making a test bed for technologies we can use in the real world.”
Researchers have built bots for the original version of StarCraft for years, using an unofficial, open-source plugin. Churchill says those bots so far are mediocre players that rely mostly on tactics coded by their designers, rather than machine learning, to build up their own grasp of the game.
Declaring war on StarCraft provides a measure of the ambition at Google and Facebook—and the limitations of today’s smartest software. StarCraft is a real-time strategy game in which players command an alien army in a distant corner of the Milky Way. While the game may appear less daunting than Go or chess, it poses a greater challenge to AI.
In chess and Go, you can see all your opponent’s moves and pieces, making them so-called perfect information games. StarCraft is an imperfect information game. You cannot see all of your opponents’ troop deployments or construction projects, forcing you to use what you’ve seen, and your mental model of the game, to predict what they may be planning.
In addition, StarCraft bots won’t be able to lean so heavily on their super-human ability to quickly crunch through myriad possibilities. The number of valid positions on a Go board is a 1 followed by 170 zeros. Researchers estimate that you’d need to add at least 100 more zeros to get into the realm of StarCraft’s complexity. “It’s a big step up,” says Oriol Vinyals, a DeepMind researcher working on StarCraft. “This game will require us to innovate in planning, memory, and how we deal with uncertainty.”
Beating StarCraft will require numerous breakthroughs. And simply pointing current machine-learning algorithms at the new tranches of past games to copy humans won’t be enough. Computers will need to develop styles of play tuned to their own strengths, for example in multi-tasking, says Martin Rooijackers, creator of leading automated StarCraft player LetaBot. “The way that a bot plays StarCraft is different from how a human plays it,” he says. After all, the Wright brothers didn’t get machines to fly by copying birds.
Churchill guesses it will be five years before a StarCraft bot can beat a human. He also notes that many experts predicted a similar timeframe for Go—right before AlphaGo burst onto the scene.