Shulou(Shulou.com)11/24 Report--

According to CTOnews.com news on July 11, Tencent announced today that Tencent AI reached the top of the international mahjong platform and reached a stable Rank of 10.68 on the Japanese mahjong Tianfeng platform, refreshing AI's best results in the field of mahjong.

▲ Tencent, according to Tencent official introduction, Tencent AI Jungyi LuckyJ went to Japan to play because "Tianfeng" is a well-known Japanese mahjong competitive platform, with more systematic competitive rules and professional Rank rules, and is widely recognized by the "professional mahjong community". Mahjong AI from all over the world basically train and play here.

Tianfeng platform ranks according to 1-11 paragraphs, of which more than 7 segments are 3037, accounting for about 1%, while only 27 people (including AI) can reach 10 paragraphs, accounting for less than 1/10000, which is equivalent to Arena of Valor's Honor King Rank.

CTOnews.com learned from Tencent officials that the top three mahjong AI are Jinji LuckyJ, Suphx and NAGA. The unique skill LuckyJ spent more than a thousand games to reach 10 stages, and all the registered players with more than 1000 games in the special room ranked first in the history of stable Rank.

It is worth mentioning that the unique skill LuckyJ not only reached the 10th paragraph of Japanese mahjong Tianfeng, but also beat six national standard mahjong professionals in the GB mahjong offline invitational tournament, becoming the first mahjong AI to beat the GB mahjong top professionals.

In order to achieve the ideal effect, Tencent AI adopts a new strategy optimization algorithm ACH (actor critic hedge), which is based on the self-game technology of "reinforcement learning and regret minimization", so that AI can "learn and improve itself" from scratch, and finally converge to the strongest mixed strategy. At the same time, based on the idea of "optimistic value estimation", the company proposes an efficient "imperfect search method", which makes it possible to combine deep reinforcement learning with imperfect information search.

Tencent said that training mahjong AI is actually training AI to better understand the human world. Finally, AI can enter into those human lives as soon as possible, such as financial transactions, autopilot, traffic logistics, auction systems and so on, to solve complex problems in the real world.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.