Google announced today that DeepMind AI is the first artificial intelligence to reach the Grandmaster level in StarCraft II: Wings of Liberty. For those who don’t know how to play the game, two players basically compete against each other after choosing which alien race to be – Zerg, Protoss and Terran – with each having different abilities. To train the AI, DeepMind used three separate neural networks – one for each race of aliens it played as. Read more for a video and additional information.
DeepMind also tapped into a large database of past games provided by Starcraft’s developer Blizzard, and then used it to train its agents to imitate the moves of the strongest players. Clones of these agents were then pit against each other to refine their skills using reinforcement learning. There were also “exploiter agents” created whose job was to expose weaknesses in the main agents’ strategies to let them find ways to correct them.
One of the key things we’re really excited about is that Starcraft raises a lot of challenges that you actually see in real-world problems. We see Starcraft as a benchmark domain to understand the science of AI, and advance in our quest to build better AI systems,” said Dave Silver, who leads the lab’s reinforcement learning research group.