AlphaStar, Google DeepMind's artificial intelligence prevails over StarCraft II players 10-1

AlphaGo is an artificial intelligence created by Google's DeepMind in 2015. This software has the ability to play the fine go and has defeated many of the world's leading goat grandmasters.

Recently, another Google DeepMind AI named AlphaStar has defeated people in a completely different game, real-time strategy game StarCraft II: Wings of Liberty.

AlphaStar began training StarCraft II: Wings of Liberty by reviewing the video of professional gamers. The time this AI learns the steps of StarCraft is equivalent to 200 years in real life. After that, AlphaStar was taught by DeepMind how to play StarCraft as the best players through competitive models. After the learning process is over, DeepMind selected 5 best versions of AlphaStar to compete with the best StarCraft II players in the world.

AlphaStar, Google DeepMind's artificial intelligence prevails over StarCraft II players 10-1 Picture 1

The match was held in December 2018 and until late last night some notable situations were broadcast on YouTube and Twitch.

The first AlphaStar competes with a player named TLO, a Zerg player in StarCraft. However, this game seems to be unfair because TLO has "had" to play Protoss for research purposes. AlphaStar results win 5 white games.

Another AlphaStar version also won 5 white games when it was played against a player named MaNa. However, on the live broadcast last night, MaNa defeated AlphaStar in a new game match.

AlphaStar demonstrated impressive ability to control troops with fast movement, good speed control in battle . AlphaStar has a lower number of action per minute than gamers but decided to make it smarter. .

However, in the play of AlphaStar, there is also a gap and MaNa discovered it. In order to crush the opponent, this AI usually moves almost the entire army towards the Maan base. Taking advantage of that, MaNa constantly sends a small group of his troops towards the back of AlphaStar's base and harassing. For defense, AlphaStar will send its entire army back to the base. MaNa takes advantage of this time to build a stronger force. And finally the player defeated AlphaStar.

AlphaStar, Google DeepMind's artificial intelligence prevails over StarCraft II players 10-1 Picture 2

To summarize, AlphaStar defeated StarCraft II with a 10-game victory and lost only 1 game. However, it is possible that in the competition, AlphaStar will win a great victory.

Compared to Go Go, StarCraft II is a more complicated game for AI. Because every time in StarCraft, AI can perform more than 300 actions including moving characters, selecting items . In addition, StarCraft possesses non-linear gameplay so when competing AI needs to build goals. for itself and must consider the effect of each operation on the winning ability of the entire game.

See more:

1. AlphaGo continues to prove its unbeaten ability in Go
2. Scientists have created the world's strongest AI, defeating the best AI in chess
3. AI has created faces like real people, faking cars, houses and animals