Subscribe to our newsletter

     By signing up, you agree to our Terms Of Use.


    • About Us
    • |
    • Contribute
    • |
    • Contact Us
    • |
    • Sitemap

    Ke Jie, Humanity’s Last Hope, Loses to AlphaGo by Half a Point

    Google DeepMind’s upgraded Go-playing computer program proves it is master of a game long thought unmasterable by AI.
    May 23, 2017#technology#sports

    The world’s top-ranked Go player, Ke Jie, has lost the first of three matches against his daunting artificial adversary, AlphaGo, and by the slimmest of margins.

    The highly anticipated contest was held Tuesday in Wuzhen, a city two hours’ drive from Shanghai, as part of the Future of Go Summit, an event jointly sponsored by the Chinese Go Association, the Zhejiang Provincial Sports Bureau, and Google. Two more games will be held on Thursday and Saturday.

    Ke, 19, stands to win $1.5 million if he bests his rival, whom he initially declined to play after AlphaGo defeated 18-time world champion Lee Sedol of South Korea four games to one in March of last year. Developed by Google DeepMind in London, AlphaGo has the ability to simulate a million matches every day, learning and improving over time at a mind-boggling pace.

    AlphaGo’s reputation apparently preceded the AI powerhouse, as mainstream platforms such as video-streaming site Youku and state broadcaster China Central Television did not broadcast the match. Hardcore Go fans, however, managed to tune into the contest through side channels for Chinese audiences that updated virtual game boards with the players’ latest moves.

    Go, or weiqi as it is known in Chinese, is a game of incalculable permutations. First played in China around 3,000 years ago, it generally involves two players taking turns placing black and white stones on a grid, with the goal being to surround the opponent’s stones and thus control more territory. Compared with the number of spaces on an 8-by-8 chess board, the number of spaces on a 19-by-19 Go board leaves little room for systematization and ample room for creativity and intuition. For this reason, Go has been regarded as sacred, human-only territory since Deep Blue, IBM’s chess-playing supercomputer, defeated Russian grandmaster Garry Kasparov in 1997.

    Even while reluctant to agree to a head-to-console match with AlphaGo, Ke, never known to be lacking in confidence, talked a big game, posting on social media, “It may defeat Lee Sedol, but not me.” Then a mysterious player with a username that translated to “Master” amassed a streak of more than 50 wins against the world’s elite players without suffering a single loss. On Jan. 4, 2017, it was revealed that “Master” was in fact an updated version of AlphaGo — and, notably, that three of those wins had come against Ke.

    After the shock, Ke seemed less sure of his chances, saying, “I don’t think I’ll be able to win easily if I end up confronting the machine.” His peers agreed, with former world champ Gu Li giving him just a 10 percent chance to win even a single game against AlphaGo — odds some still considered generous. Nevertheless, Ke agreed in April to face AlphaGo this week in Wuzhen.

    According to the rules of the contest, Ke and AlphaGo were given three hours over which to play their moves. By the time the match ended, Ke, staring intently at the board with the palm of his hand propped against his face, had just 13 minutes of time remaining. AlphaGo, meanwhile, had used just under half of its allotted time, sitting comfortably at 1 hour, 29 minutes. When both players finally declined to place any more stones, neither had a visible advantage. After the final count, Ke Jie learned that he had lost by half of one point — the smallest possible margin.

    Online, some Chinese netizens saw Ke’s defeat as a fitting opportunity to remind him of his overconfidence. Wang Sicong, the outspoken scion of the Wanda real estate empire, weighed in on microblog platform Weibo, writing in a post that has since been deleted, “Where is the cocky you we saw when Lee Sedol was defeated by AlphaGo?” Another user quipped, “Ke Jie is still the greatest person I’ve ever seen when it comes to bragging.”

    By and large, however, net users have rallied with support for Ke, calling him “a hero” regardless of the outcome. And Demis Hassabis, DeepMind’s co-founder, said during the press conference following the match that he had “huge respect [for] Ke Jie for the amazing, brave game he fought today.”

    Facing the crowd, Ke, gracious in defeat, conceded that AlphaGo was “really good” and called the opportunity to play against such a sophisticated AI “the greatest honor” of his life. “Last year, AlphaGo played more like a human,” he said, “but right now, it’s playing more like a god of Go.”

    Contributions: Qian Zhecheng; editor: Kevin Schoenmakers.

    (Header image: The world’s No. 1 Go player, 19-year-old Ke Jie of China, considers his next move during the first of three matches against Google’s AI computer program, AlphaGo, at the Future of Go Summit in Wuzhen, Zhejiang province, May 23, 2017. IC)