Today's "Man-Machine War" | Alphago Master first innings beat Coger _ depth study

May 23, "China Go Summit" in Wu Town, the world's first chess player Coger and Alphago Master's first game began at 10:30, 14:50, three chess game first, Alphago White 1/4 son wins, the score 0-1. Alphago at present in the strength has had the more obvious superiority, basically controls the entire game situation, has defeated the Coger smoothly. The new version Alphago abandoned the human games, only by the way of depth learning to challenge the limit of Weiqi, live address: http://events.google.com/alphago2017/.

The overall schedule is as follows:

May 23 Coger VS Alphago
May 24 Artificial Intelligence Summit Forum
May 25 Coger VS Alphago
May 26 +alphago VS Laugh +alphago | When the more + Chen Yaoye + 芈 Yu ting + Tangweising + Zhou (talk Chess) VS Alphago
May 27 Coger VS Alphago

Coger and Alphago's battle scene.

DeepMind founder Demis Hassabis, Coger and Google chairman Eric Emerson Schmidt

On the eve of the opener, Coger update microblogging said "Now the AI progress far beyond our imagination, but it is always a cold machine, compared with the human, I do not feel it on the passion and love of go, regardless of the result, this will be my last three machine war." About Alphago

What Alphago is.

Alphago is the first to beat the human professional Go player and win the Go World Championship program, is one of the most powerful players in Weiqi history. March 2016, in the world more than 100 million viewers of the attention, Alphago after 5 innings, and finally to 4:1 of the total score defeated the go World champion Li Shishi, this game has become an important milestone in the field of artificial intelligence. In the past, experts have predicted that artificial intelligence would take 10 years to defeat a human professional. After this game, alphago by virtue of its "full of creative and witty" dismount, into the go world's highest professional title-career nine ranks, become the history of the first to obtain this honor of the non-human chess player.

Recently, Alphago's upgraded version of the "Master/magister" title with the world's top Weiqi players carried out 60 online fast chess game, and achieved a good record of victory.

Alphago how to train.

Weiqi has long been considered one of the most challenging projects in traditional games for artificial intelligence. This is not only because Weiqi contains a lot of search space, but also because the evaluation of drop position is far more difficult than the simple heuristic algorithm.

In order to deal with the great complexity of Weiqi, Alphago adopts a novel machine learning technology, which combines the advantages of supervising learning and strengthening learning. By training to form a strategy network (Policy Network), the situation on the chessboard is used as input information, and a probability distribution is generated for all feasible drop positions. Then, a value network (network) is trained to predict the results of all feasible drop positions by the criteria of 1 (absolute victory of the opponent) to 1 (absolute victory of Alphago). Both networks are very powerful themselves, and alphago the two networks into a probabilistic Monte Carlo search (MCTS) to achieve its real advantage. Finally, the new version of Alphago produces a lot of self chess chess, which provides training data for the next generation, and the process repeats itself.

Alphago how to decide drop.

After getting the chess information, Alphago will explore which location is high potential value and high probability according to the strategy network, and then decide the best drop position. At the end of the assigned search time, the most frequently inspected location in the simulation process will become the final choice of Alphago. After the thorough exploration and the process of the best drop, the Alphago search algorithm can add approximate human intuitive judgment to its computational ability.

Alphago's Chifeng

The most powerful place in Alphago is not embodied in a particular piece of chess or a local change, but in the unique perspective it presents in every game of chess. Although Alphago's Chifeng is not easily summed up in itself, Alphago is more likely to use a free, open style of chess in general. In its world, there is no preconceived notions, and there is no rule that must be obeyed, which allows it to break the rules and find the most efficient hand in the current chess game.