Google’s artificial intelligence lab DeepMind announced Thursday that it has successfully beaten two professional StarCraft II players with AI.
Happy that we could share #AlphaStar progress with you all! Good Games @LiquidTLO and @Liquid_MaNa, and @Artosis and @RotterdaM08 for a great show! You can see all the details in the blog.https://t.co/ZAOMsVx8pY pic.twitter.com/51EG3fHUL1
— Oriol Vinyals (@OriolVinyalsML) January 24, 2019
In preparation for the matches, DeepMind’s specially trained AI “AlphaStar” was fed 200 years’ worth of StarCraft II playing experience.
Facing off against Team Liquid’s Grzegorz “MaNa” Komincz and Dario “TLO” Wünsch, AlphaStar was able to beat both players over the course of five matches each.
Wünsch even described the AI as playing “completely different” than any human he has encountered.
While AI has been successful in popular games such as Quake III arena and Dota 2, AlphaStar’s victory is significant given the complexity of StarCraft II, arguably one of the most challenging real-time strategy games ever made.
As noted by Kotaku, AlphaStar was able to win not by making decisions faster but by making better and more thought-out decisions in general. In fact, the number of clicks and key presses made per minute by the AI was “significantly lower than the professional players.”
DeepMind also revealed that AlphaStar opted to view the game’s entire map at once as opposed to being zoomed-in and focused on specific parts like its human opponents.
“Human players must explicitly manage an ‘economy of attention’ to decide where to focus the camera,” DeepMind said. “However, analysis of AlphaStar’s games suggests that it manages an implicit focus of attention.”
Although AlphaStar’s accomplishment is noteworthy in the gaming world, Demis Hassabis, CEO and cofounder of DeepMind, argues that the technology can potentially be just as important for more real-world applications.
In remarks on Twitter Thursday, Hassabis said that the brains behind AlphaStar “could be useful in other problems such as weather prediction & climate modeling.”
3/3 While StarCraft is ‘just’ a (very complex!) game, I’m excited that the techniques behind #AlphaStar could be useful in other problems such as weather prediction & climate modeling, which also involve predictions over very long sequences. Peer-reviewed paper is underway.
— Demis Hassabis (@demishassabis) January 24, 2019
- Elon Musk’s Starship prototype has literally blown over
- John McAfee to run 2020 bid while exiled on a boat
- Hacker infiltrates Nest cameras to gain PewDiePie subscribers