Expert: AlphaGo in depth learning strength has not yet reached top | | AI Li Shishi | AlphaGo_ news

Institute of automation Chinese Academy of Sciences, complex systems management and control of national key laboratory of Wang Fei jumping trip to Beijing, video interviews. Beijing, reporter Fu Yu photography

Beijing, March 15, artificial intelligence AlphaGo and Korea five war player Li Shishi is currently a hot topic. This afternoon, Chinese Academy of Sciences Institute of automation of complex systems management and control of national key laboratory of Wang Fei jumping trip to Beijing, (micro public number: cns2012), "the man-machine war" in the focus of many about his views.

The King leaps, AlphaGo was able to defeat Li Shishi was winning in its depth of learning and search capabilities.

"Go is difficult to overcome, difficult lies in the changes so much. If you AlphaGo a hard count, listing all the possibilities, and then select the best should start, that's not possible. AlphaGo used his network of complex space has shrunk to such a range can be calculated, this is in fact the realization of deep learning. And AlphaGo according to the probability of further optimization and evaluation methods, and ultimately choose the best begin, which is its biggest breakthrough. ”

While King leap also believes that AlphaGo strength is far from the pinnacle of artificial intelligence.

"Defeating Li Shishi, AlphaGo proved by go the power, but only mankind has invented artificial intelligence learning rules and algorithms. Prior to this, artificial intelligence has been breakthrough in speech recognition, image recognition, and other fields. In addition to deep learning, there are many methods of artificial intelligence, I don't think it reached a peak of artificial intelligence. ”

Responsible editor: Chen yan SN225

Article keywords:
AI Li Shishi AlphaGo

I want feedback
Save a Web page
China News Network

专家：AlphaGo强在深度学习 实力还未达巅峰|人工智能|李世石|AlphaGo_新闻资讯

　　中国科学院自动化研究所复杂系统管理与控制国家重点实验室主任王飞跃做客中新网视频访谈。中新网记者 富宇 摄

　　中新网3月15日电 人工智能AlphaGo与韩国棋手李世石的五番战是时下最火热的话题。今天下午，中国科学院自动化研究所复杂系统管理与控制国家重点实验室主任王飞跃做客中新网(微信公众号：cns2012)，针对“人机大战”中的诸多焦点谈了自己的看法。

　　在王飞跃看来，AlphaGo之所以能击败李世石，是胜在它的深度学习和搜索计算能力。

　　“围棋之所以难被攻克，难就难在它的变化非常多。如果AlphaGo硬算，把所有的可能性罗列出来，然后从中选取最佳应手，这是不可能的。AlphaGo利用他的网络，把这种复杂的空间收缩到一个可计算的范围内，这其实就是深度学习的实现。然后AlphaGo再根据概率和评估方法做进一步优化，最终选择出最佳应手，这是它最大的突破所在。”

　　同时王飞跃还认为，AlphaGo的实力还远未达到人工智能的巅峰。

　　“击败李世石，AlphaGo通过围棋证明了它的威力，但也只是说明了人工智能学习了人类已经发明的规则和算法。在这之前，人工智能已经在语音识别、图象识别等领域有所突破。除深度学习之外，人工智能还有许许多多的方法，我不认为它达到了人工智能的高峰。”

责任编辑：陈琰 SN225

文章关键词：
人工智能 李世石 AlphaGo

我要反馈
保存网页
中国新闻网