Human players are no match at video games for DeepMind
The artificial intelligence crushes its human rivals at StarCraft 2
AlphaStar is practically invincible. DeepMind’s artificial intelligence (AI) I can now crush almost every human player in StarCraft II. This company became truly notable at mastering difficult games such as chess or shogi and Go and DeepMind’s AI is now smashing human brains with their machine learning proficiencies.
This same year AlphaStar was released, this is an Al built real-time strategy game for StarCraft II. It was revealed by DeepMind with the already tradition of embarrassing human players, stomping on the best performers on StarCraft II. Last Wednesday, the journal Nature published a study by the DeepMind team that covers the details of AlphaStar’s performance. At this point, it is now classified as Grandmaster and, bad news for the new rising stars, it can beat 99.8% of all human players.
To speak about the process that follows getting a machine to a success point, it starts with AlphaStar watching a million replays of different games and strategies for the machines to learn. Once there, the AI confrontes top players to prove the strategies selected are right or wrong. Then, it involves a period of teaching the machines that winning is good and losing is wrong so they can aim to win.
Thinking about the reasons that will lead a company to invest time and efforts on an artificial intelligence video game player, development comes to mind. The intention behind DeepMind is to contribute to developing machines to improve real-life routines. Conquering a video game that is also played live, is an important step to understand better this technology and transport this to the future, with vehicles driven by machines.