DeepMind’s AlphaStar AI to Play Blind StarCraft II Matches Against Regular Human Players

AlphaStar, an AI created by Google-owned DeepMind that plays StarCraft II, recently flexed its gaming muscles by absolutely destroying two professional human players in the strategy game. For those who think they can fare better than the two defeated Team Liquid players against AlphaStar, they now have an opportunity to prove it. Blizzard is offering players a chance to play against DeepMind's AI in blind matches to prove their mettle. The initiative has been launched as part of ongoing research by DeepMind that will assess AlphaStar's performance for scientific purpose.

In an official blog post, Blizzard has announced that AlphaStar will play a series of 1v1 matches on a blind trial basis in Europe against human players. For StarCraft II players who want a stab at AlphaStar, they can tap on the “DeepMind opt-in” button on the 1v1 Versus menu to get a chance and play against the AI. However, players won't know if they are playing against AlphaStar or a human opponent. The reason behind it is to rate the AI's performance against general gaming behaviour of regular players.

Experimental versions of AlphaStar will be pitted against players via the regular matchmaking rules, however, DeepMind has not revealed the frequency at which players will get to play against the AI. It is being done to ensure anonymity so that the choices made by StarCraft II players and their overall game strategy do not change on discovering that they are playing against the AI.

AlphaStar will play as a member of Terran, Zerg, or Protoss races in the game, and will only be tested against StarCraft II players in Europe. All the rules will be normal, be it match-making or victory points, and players won't win any additional rewards in case they defeat the AlphaStar.

Heck, they won't even know that they've beaten the expert AI, which easily defeated two professional players from one of the world's best esports teams. DeepMind, on the other hand, will release the performance metrics of AlphaStar during the blind matches against human players in a peer-reviewed scientific paper after the test has concluded.