On January 25, the result of officially playing of the artificial intelligence “AlphaStar” developed by DeepMind of Google against the pro gamer of RTS “Starcraft 2” was announced. DeepMind’s AI won against the StarCraft II professional player.
On a press conference, DeepMind, the Google branch dedicated to artificial intelligence, showed the improvement of its work on the video game “StarCraft II”. According to the progress, even in games AI can win against humans.
Deep Reinforcement Learning of “StarCraft II” being studied through collaboration between Google’s affiliated DeepMind Technologies and Blizzard Entertainment (“Blizzard”). As one result, it has been reported that AI won the human professional player in the battle mode in the same work.
“No system has come anywhere close to rivaling the skill of professional players. In contrast, AlphaStar plays the full game of StarCraft II, using a deep neural network that is trained directly from raw game data by supervised learning and reinforcement learning,” Google said.
Back in 2017, AI AlphaGo won the top players at Go, then DeepMind was preparing a setup “AlphaStar” to play more complex real time strategy games.
Team Liquid’s TLO and MaNa are the pro gamers, who fought against AlphaStar, but AlphaStar won 5 wins and 0 losses for TLO players, 5 wins and 1 loss for MaNa players, 10 wins and 1 loss in overall performance.
On the AlphaStar side, the response speed and the APM (manipulated amount per minute) are suppressed to the same level as human in order to make sure fair match conditions as possible.
TLO players who battled said:
“It is very difficult to compete with AlphaStar, it is very difficult to compete against AliStar, not playing against AI, it’s a play that I have never experienced before, AlphaStar is truly amazing, games I think that there is no unprecedented artificial intelligence.”
Now, it has been proved this time that the selection is greatly more than Shogi and Go and the RTS game where the situation changes constantly and the artificial intelligence level as good as to the top players of human beings.