Facebook’s artificially intelligent Go player is getting smarter
Facebook today is demonstrating recent progress in its effort to have a computer program beat a human being in the ancient Chinese board game Go.
In an academic paper that Facebook published online today, Facebook documents the advancements of its Darkforest Go players that draw on artificial intelligence (AI) in order to more effectively win games.
“Against human players, the newest versions, darkfores2, achieve a stable 3d level on KGS Go Server as a ranked bot, a substantial improvement upon the estimated 4k-5k ranks for DCNN reported in Clark & Storkey (2015) based on games against other machine players,” Facebook AI researchers Yuandong Tian and Yan Zhu wrote in the paper’s abstract. “Adding MCTS [Monte Carlo Tree Search] to darkfores2 creates a much stronger player named darkfmcts3: with 5,000 rollouts, it beats Pachi with 10k rollouts in all 250 games; with 75k rollouts it achieves a stable 5d level in KGS server, on par with state-of-the-art Go AIs (e.g., Zen, DolBaram, CrazyStone); with 110k rollouts, it won the 3rd place in January KGS Go Tournament.”
「KGSの囲碁サーバーの囲碁のランキングボットで安定的に3dレベルを達成している最新バージョンのdarkfores2は、他の機械のプレイヤーに基づいたClark & Storkey (2015年版) で、DCNNのための4kから5kランクと見積もられるほど、かなりの改善が報告された。」 Facebookの人工知能研究者である Yuandong Tian と Yan Zhu が、論文のアブストラクトに書きました。 「MCTS（モンテカルロ木探索）をdarkfores2へ加えることは、darkfmcts3と名づけられた、より強いプレイヤーを作り出した。darkfmcts3は、1万回のロールアウトの Pachi との250回の対戦の全てで、5000回のロールアウトで勝った。7万5千回のロールアウトで、KGSサーバーの安定的な5dレベルを達成。最先端の囲碁のAI（たとえば、ZenやDolBaram, CrazyStoneなど）と同等。1月のKGSトーナメントでは、11万回のロールアウトで3位を勝ち取った」
- rollouts - ごめんなさい、この囲碁の文脈での「ロールアウト」の意味が分かりません。「手」？「先読み」？
Facebook has been trying to make computers successfully beat humans in Go in just the past few months. But Go has actually been an area of focus for researchers for decades.
Facebook has been rapidly building up AI talent and technology, and it has open-sourced software and shared the designs of its AI hardware.
In recent years IBM has been the company to impress people with its ability to beat human beings with technology at certain specific challenges — namely Watson beating human beings on Jeopardy and, earlier, Deep Blue beating Garry Kasparov in chess. But the technology that Facebook is employing here is
Watch an AI bot play Go
To help teach systems how to plan, Facebook's AI research team has created an AI bot to play the board game Go. After a few months of playing, it's already on par with the other AI-powered systems that have been published and it's as good as a very strong human player. We've achieved this by combining the traditional search-based approach — modeling out each possible move as the game progresses — with a pattern-matching system built by our computer vision team.Posted by Facebook Engineering on 2015年11月3日