After Go, Google’s Deepmind plans to take on Starcraft II

After having proved its might against humans by beating the world champion Lee Sodol at one of the toughest games in human history – Go, Google's Deepmind AI is eyeing a new territory of strategy- based game Starcraft II.

Moving forward, Google has teamed up with Blizzard to extend the scope of its neural networks to widely popular game Starcraft II and will eventually open the game as a research platform for those building AI programmers.

"Blizzard will release an API early next year that will allow researchers and hobbyists around the world to build and train their own AI agents to play Starcraft II," said Oriol Vinyals, a research scientist at Google DeepMind.

“StarCraft is an interesting testing environment for current AI research because it provides a useful bridge to the messiness of the real world. The skills required for an agent to progress through the environment and play StarCraft well could ultimately transfer to real-world tasks,” he added.

Notably, the project isn’t just an AI to play StarCraft II, the idea instead is to create an AI-friendly environment for the game, something that integrates the features of AIs to access in order to understand this complex and visually messy game.

For the game of Go, DeepMind learned by teaching itself through trial and error. All the researchers did was to explain how to determine success, and the AI could then begin playing games against itself on a loop while always reinforcing any strategies that lead to more success.

For StarCraft, that will likely mean asking the AI to prioritize how long it survives and/or how much damage it does to the enemy’s primary base. As they play and watch recorded games, the AI will build up knowledge about strategy, tactics, the so-called “macro” game, while learning how best to employ its superhuman actions-per-minute rate to dominate the “micro” in battles and skirmishes.

As TechCrunch puts it, “It’s not just for the glory of the win (though that’s part of it) — the creation of an AI that can handle this level of visual and gameplay complexity will aid in understanding what our models have trouble with and what they excel in.”