Man-machine chess wars heated up, Kai-Fu Lee said lishishixin play not well-Li Shishi, AlphaGo, go-IT information

IT information news on March 9, the artificial intelligence and human-computer world war matters as much as humans finally opened, its the first game has been going on for 100 minutes ( live video address ), the race enters stage with impatience. Innovation works, Kai-Fu Lee for game analysis, said Li Shishi seemed to have tried some new way of playing, but this is not a good idea.

Kai-Fu Lee said, because the machines search is complete without games there have been moves will let it go astray this situation. Li Shishi, a good game is based on the classic standards, experts play in the past, be more sensible than desperation.

Kai-Fu Lee, who said this move WINS or loses is not the most important. Once after the AlphaGo up to the professional level of the second, beat only thing sooner or later.

Mr Lee points out that even beat Li Shishi AlphaGo, it can't tell why. It is a system of basic reasoning, logical system. Deep talk anymore, it can't approach not only explain, no feelings, no emotions. This machine is very early. The next milestone in the field of artificial intelligence, we should look at the right side of the brain of the machine should have upgraded, rather than logic should have upgraded.

It is understood that the AlphaGo use of deep learning, has two brains, he and Fan Hui game is 30 million game, is now more than 100 million. In such a case, it will make a choice, which fell chess winning even more.

人机围棋大战白热化，李开复称李世石新打法不周密 - 李世石,AlphaGo,围棋 - IT资讯

IT资讯讯 3月9日消息，对人工智能和人类而言都意义非凡的人机世纪大战终于开打，其第一局比赛已经持续了100分钟（视频直播地址），赛事也进入焦灼阶段。创新工场李开复对棋局分析称，李世石好像有在尝试一些新的打法，但这是不明智的。

李开复表示，因为机器的搜索非常完整，不会出现棋谱没有出现过的招数就会让它误入歧途这种情况。对李世石而言，一个比较好的打法还是按照经典标准的，过去专家的打法来，会比铤而走险更明智一些。

李开复先生表示其实这次棋谁赢谁输不是最重要的。一旦AlphaGo达到专业二段水平以后，打败人类只是迟早的事情。

李开复先生也提出，即便是AlphaGo打败了李世石，它讲不出为什么。它是一个基本推理的系统，逻辑性的系统。再往深讲，它不但没法办法解释，更没有喜怒哀乐，没有七情六欲。在此方面机器还是非常初期的。人工智能领域的下一个里程碑，应该看看机器的右脑方面有没有提升，而不是逻辑方面有没有提升。

据了解，AlphaGo运用了深度学习，有两个大脑，他跟樊麾比赛的是3000万盘棋，现在是1亿以上。在这样的情况下它会做出一个选择，哪种落棋胜率更大。