Google AlphaGo waste the world champion 4 to 1 – Tom’s Hardware

AlphaGo beat the world champion of Go with a final score of 4 to 1 . It is the challenge in five matches between the DeepMind supercomputer, owned by Google, and Lee Sedol, world champion in charge of Go, a board game in which the complex is used a good deal of strategy is closed this morning (Italian time) to win.

 
 
 
 

After the first three-game winning streak, the fourth round (which was played Sunday, March 13) went in favor of Lee Sedol, who managed to push AlphaGo error. The last shot went again in favor of the computer, which eventually proved to be a more than worthy opponent for sample, that the first game had admitted that he had underestimated the opponent.

 
 

The ability to AlphaGo however, were clear even before this morning, so much so that on Sunday, despite the defeat, the Korea Baduk Association had recognized the honorary evaluation of “the ninth dan”, ie the same level that has earned Lee on the field.

 
 

last shooting Lee Sedol has surrendered, giving the victory to AlphaGo , as happened in the first round because the AI ​​had become an irretrievable advantage. But this time the game has been more fought because the supercomputer has made a serious mistake similar to the one that cost him the fourth challenge. Unlike Sunday, however, the computer continued with a game almost no errors to victory.

 
 
 
 

Demis Hassabis, founder of DeepMind, in a long interview with The Verge had explained that before he races and his team expected to be on par with the enemy in the flesh . Apparently it went better than expected.

 
 

Hassabis explained that the difficulty of games like this is that they require a high level of strategic capacity in a world of imperfect information. In Go, however, the pieces are all on the table so to computers is slightly easier than other strategy games like StarCraft, a RTS (Real Time Strategy).

 
 

Hassabis reminds us that “the purpose of DeepMind is not just to beat the games [...], which are useful as a test bed. [...] The goal is to apply this knowledge to the great real-world problems.”