Lee Sedol lost to AlphaGo in the fifth and final game in th

AI Google AlphaGo won in the fifth, the final game of go, defeating one of the best human players, Lee Sedol. This time the champion was playing black stones, and the program was white. The Korean go player tried to constantly attack the enemy, as in the fourth , victorious game for himself. Then Sedol tried not to give initiative to his virtual rival.

Like last time, the time of the human player ended faster than that of the car, so Sedol began to use extra time, beat . He still managed to somewhat disturb the AlphaGo protection system, but the AI ​​solved this problem and returned the conquered territories. Almost until the very end, the players were on equal footing, but Sedol still had to admit defeat. It happened after five hours of intense play.


')
According to Google, who organized this game, if Cedol won a computer in most games, he would receive $ 1,000,000. But since the victory was left for AlphaGo, the million went to charity.

#AlphaGo wins game 5 taking the historic match 4-1. Congrats DM team and Lee Sedol. Awesome displays of human ingenuity all round!

- Mustafa Suleyman (@mustafasuleymn) March 15, 2016

The Go program itself was created by DeepMind, bought by Google some time ago. According to the creators of the program, only during the training she processed about 30 million different combinations, gradually learning various methods and schemes of the game. The calculation of optimal combinations is also used, but not for the game as a whole, but only for solving frequently arising local moments.

If you suddenly missed the broadcast, here's a video with the analysis of "flights":



Well, and a link to the file with the moves, if you want to understand in detail.

Let me remind you that Lee Sedol lost the first , second , third and fifth installments of the car. He won the fourth game, as already mentioned above.