A Well-Trained AI Took on StarCraft II Pros
Google’s DeepMind AI, which is named “AlphaStar,” just beat Team Liquid’s Grzegorz “MaNa” Komincz and Dario “TLO” Wünsch five games to nothing in two separate five-game StarCraft II series. Prior to the games, Wünsch watched footage of AlphaStar playing and said he was “extremely confident” going in. But after being hit rapid-fire with different strategies, he was unable to read his digital foe.
One reason for AlphStar’s success is that it played each game as a different “agent” with its own unique strategy. In addition, it “sees” StarCraft in a different way than human players by using an interface that gets a top-down view of the whole map as opposed to small views that must be zoomed-in at different points.
“Human players must explicitly manage an ‘economy of attention’ to decide where to focus the camera,” the DeepMind team said. “However, analysis of AlphaStar’s games suggests that it manages an implicit focus of attention. On average, agents ‘switched context’ about 30 times per minute.”
Komincz said that playing the AI gave him a new perspective on StarCraft II, which he previously thought he knew inside and out.
“I was impressed to see AlphaStar pull off advanced moves and different strategies across almost every game, using a very human style of gameplay I wouldn’t have expected,” he said. “I’ve realized how much my gameplay relies on forcing mistakes and being able to exploit human reactions, so this has put the game in a whole new light for me.”