Li Shishi field changed: 5:0 win a computer, now it's hard to say-Li Shishi, AlphaGO, go-IT information
March 9, Google DeepMind team to develop artificial intelligence go program AlphaGo and Korea around player Li Shishi started five games of the contest. March 8, at 10 o'clock in the morning, the "man-machine war" first media conference in Korea Seoul Seasons Hotel 6-floor conference room.
Start from 8:30 A.M., the media came on the scene, when began the press conference is still an hour away, launches have full machine in the back, the first few rows of seats have been filled. Small Banquet Hall finally squeezed into the more than 400 media, 80% of them were Korea local outlets.
Google Chairman Eric Schmidt as a special guest came to the scene, said he has always "I don't know how the outcome of the game, but, in any case, was a victory for humanity. ”
AlphaGo Demis Hassabis show live media development leaders on the algorithms of artificial intelligence programs, emphasizing it with Zen, Crazystone of which Kyu, the biggest difference is that it is not "death count", but rather "selective searches" and choice with artificial intelligence like IBM deep blue "predecessors" compared to Demis AlphaGo was the first universal artificial intelligence (Artificial General Intelligence,AIG), compared with only for programs that perform a single task, AlphaGo can solve the more general problem, he's most likely to take the lead in areas of application are: medical, robotics and personal assistant.
But the whole show was most continued to be in advance of the war hero, Korea player Li Shishi. "When Google found me, I was surprised. I thought the level of artificial intelligence to achieve human top player, should take ten years, did not expect to be so fast. "Li Shishi said at the press conference," but I am also very curious, and curiosity is the best way I have personally and it matches, so consider a 3 seconds I said yes. ”
Despite promises fast, but Li Shishi also admits that before him all through the game is different, "rivals is a AI, so I need to use a special solution". He did not think AlphaGo before the data reference and Fan Hui, because at that time the AlphaGo did not reach the level of his level.
"Though five months have passed, and I don't know how much it has progressed. Especially after I learned about its algorithm, I feel I might not win 5:0 "Li Shishi says," because I know it works after the discovery, although I rarely make mistakes, but I made a small mistake, you might actually lose. ”
李世石现场改口：要5:0赢电脑，现在很难说 - 李世石,AlphaGO,围棋 - IT资讯
3月9日开始，Google DeepMind团队开发的人工智能围棋程序AlphaGo将与韩国围棋手李世石展开五场比赛的较量。3月8日上午10点，这次“人机大战”的第一场媒体发布会在韩国首尔的四季酒店6层的会议厅举行。
从早上8点半开始，就有媒体陆续来到了现场，在离发布会开始还有一个小时时，发布会后排已经架满了机器，前几排的坐位也已经被占满。不大的宴会厅里最后挤进了超过400家媒体，其中80%以上是韩国本土媒体。
Google公司董事长埃里克·施密特也做为特别嘉宾来到了现场，他以一贯的口吻说“我不知道比赛的结果如何，但不管如何，都是人类的胜利。”
AlphaGo的开发领导人Demis Hassabis向现场媒体展示了这台人工智能程序的算法原理，强调它与此前Zen、Crazystone这些围棋程序最大的不同是它并不是“死算”，而是“有选择性的搜索”和选择；而与IBM 深蓝这样的人工智能“前辈”相比，Demis觉得AlphaGo是第一台通用人工智能（Artificial General Intelligence，AIG），与那些只为执行单一任务的程序相比，AlphaGo能解决更多普适性的问题，他给出的最有可能率先应用的领域分别是：医疗、智能机器人和私人助理。
不过整场发布会被提前最多的仍然是这次大战的主角，韩国棋手李世石。“当Google找到我时，我还挺惊讶的。我以为人工智能要达到人类顶尖棋手水平，应该还要十年时间，没有想到会这么快。“李世石在发布会上说，“不过同时我也很好奇，而解决好奇最好的办法就是我亲自和它比赛，所以考虑了3秒钟我就答应了。”
尽管答应得很快，但李世石也坦承这和他之前所有经历的比赛都不一样，“对手是一个人工智能，所以我也需要用一些特别的办法准备”。他不认为之前AlphaGo与樊麾较量的数据有参考意义，因为那个时候的AlphaGo完全没有达到他这个级别的水准。
“不过已经过去了五个月，我不知道它进步了多少。尤其是我了解了它的算法之后，我觉得我有可能不会5:0轻松获胜”，李世石说，“因为我了解了它的原理之后发现，虽然我很少犯错，但如果我犯了一个小错误，就有可能真的会输。”