Google AI Defeats Human Go Champion
- 2017年5月26日
- 読了時間: 5分
グーグルの人工知能プログラムDeepMind AlphaGoが、囲碁のチャンピオンに勝った、というBBCのニュース記事。後に関連動画とディクテーションタスク、スクリプトを付けました。
artificial Intelligence: 人工知能(AI)
secure a victory: 勝利を確実なものにする
push . . . to the limit: ～を限界に達せさせる
a bit of a: ちょっとした
take turns: 交代で行う
grid: 格子[碁盤目]状(のもの)
deploy: (軍隊などを)配備・展開する、(布陣を)敷く
prof = professor
algorithm [ǽlgərìðm]: アルゴリズム(数学的な問題を解くための手順)
govern [ǽlgərìðm]: ～を統治する、～を管理・運営・統制する
diagnosis [ǽlgərìðm]: (病気の)診断
その他わからない語句は英辞郎で:
こちらはロイター通信によるニュース動画(1分18秒)。英語は中国語訛りのある米国英語です。
【スクリプト】
Another victory for artificial intelligence over the human brain. The grand champion of one of China’s most ancient and iconic board games, throwing a towel on Thursday in a man-versus-machine grudge match against an AI system powered by Google. It’s the second victory over champion Ke Jie this week for AlphaGo, *making their final meeting in the best-of-three competition on Saturday essentially meaningless.
In their first game, AlphaGo defeated the human master by just half a point, but Thursday’s victory was far more emphatic, **Ke Jie quitting before the game was officially over. After the game, he said playing with AlphaGo gave him (a) new understanding of the game, even (a) new understanding of life. He is now the second champion to bow to AlphaGo’s supremacy, after the Korean number one was defeated in four out of five games by the program last year.
That contest got (a) lot of publicity in China, but with the competition on home soil this year, there is a lot less online buzz. Games have been officially streamed live on YouTube, which is owned by Google. But YouTube has been blocked in China for years along with all other Google platforms after the company refused to censor its search results.
iconic: 象徴的な
throw a towel: 降参する
grudge match: 遺恨試合
emphatic: 強調された、明らかな、断固たる
*make . . . meaningless:〈使役〉～を無意味なものにする
best of three competition:〈囲碁〉三番勝負
*Ke Jie quitting . . .:〈独立分詞構文〉→BT246
bow to: ～にお礼[おじぎ]をする、～に降参する
supremacy [supréməsi ]: 優位
on (one’s) home soil: (自分の)国の領域内で
buzz: 蜂などのブンブンいう音、(会話の)ざわめき
censor: 検閲する
こちらはDeepMindの創設者Demis Hassabisのインタビュー動画(2分47秒)。英語はイングランド英語です。
【ディクテーション】
Games are (kind of) (a) microcosm of the outside world. That’s why games were invented, (and) that’s why humans find it fun to play.
[ ] [ ] [r ] [h ] of computer-tackling board games (that) started with games like backgammon, drafts, and then finally there was Deep Blue in ’97 that beat Kasparov at chess, which was obviously a huge watershed moment for game AI. Since then, the really big remaining (sort of) holy grail, if you like, has been Go.
In chess, the number of possible moves is about twenty for the average position. And (in) Go, it’s about 200. Another way of [v ] [ ] [c ] of Go is that the number of possible configurations of the board is more than [ ] [ ] [ ] [a ] [ ] [ ] [u ].
If you ask a great Go player why they played a particular move, sometimes they will tell you [ ] [f ] [r ]. So you can . . . the one way you could think of it is that Go is a much more [i ] game, whereas chess is a much more [l ]-[b ] game.
The AlphaGo is our program to actually try and crack Go. We played the European champion Fan Hui and a five-game mach. We have an estimate of how strong a program is. And of course you never know when you play a human player, they do all sorts of very interesting, creative things that your program does not necessarily do, so there’s always an unknown.
I think he was a bit stunned after the first game, and I think he was unsure as to whether he [ ] [p ] [e ] enough or whether the program was really strong.
AlphaGo won 5-0(five-nil), which is the first time ever a program has beaten a professional player.
The way I see where we are now is that we’ve beaten the European champion now, and the next step for us is to try and challenge the legendary player Lee Se-dol, who you could [t ] [ ] [a ] [ ] [ ] [F ] of Go.
We always used to talk about, well, if we could eventually be able to crack Go and have a program that could beat the world champion, then we must have invented some generic, general-purpose algorithm, so maybe we’re on the cusp of all of that. And we’re very [e ] [ ] [ ], but it is just one rung on the ladder towards solving artificial intelligence.
【スクリプト】
Games are (kind of) (a) microcosm of the outside world. That’s why games were invented, (and) that’s why humans find it fun to play.
There’s a rich history of computer-tackling board games (that) started with games like backgammon, drafts, and then finally there was Deep Blue in ’97 that beat Kasparov at chess, which was obviously a huge watershed moment for game AI. Since then, the really big remaining (sort of) holy grail, if you like, has been Go.
In chess, the number of possible moves is about twenty for the average position. And (in) Go, it’s about 200. Another way of viewing the complexity of Go is that the number of possible configurations of the board is more than the number of atoms in the universe.
If you ask a great Go player why they played a particular move, sometimes they will tell you it felt right. So you can . . . the one way you could think of it is that Go is a much more intuitive game, whereas chess is a much more logic-based game.
The AlphaGo is our program to actually try and crack Go. We played the European champion Fan Hui and a five-game mach. We have an estimate of how strong a program is. And of course you never know when you play a human player, they do all sorts of very interesting, creative things that your program does not necessarily do, so there’s always an unknown.
I think he was a bit stunned after the first game, and I think he was unsure as to whether he hadn’t prepared enough or whether the program was really strong.
AlphaGo won 5-0(five-nil), which is the first time ever a program has beaten a professional player.
The way I see where we are now is that we’ve beaten the European champion now, and the next step for us is to try and challenge the legendary player Lee Se-dol, who you could think of as the Roger Federer of Go.
We always used to talk about, well, if we could eventually be able to crack Go and have a program that could beat the world champion, then we must have invented some generic, general-purpose algorithm, so maybe we’re on the cusp of all of that. And we’re very excited about it, but it is just one rung on the ladder towards solving artificial intelligence.
microcosm: 小宇宙、縮図
watershed moment: 重大な分岐点
holy grail: 困難な探求の対象、至高の目標
if you like: 言ってみれば
stunned: 唖然・愕然・呆然として
as to: ～に関しては
nil: (主に英) (スポーツの試合の)ゼロ点
generic: 市販の
cusp: 先端、(新しい時代の)幕開け、入口
rung: (はしごなどの)横木、段、段階
Wikipedia - Demis Hassabis:
Comments