论文部分内容阅读
麻省理工学院技术评论网站2016年12月15日刊登文章称,越来越多的企业在应用先进的人工智能技术,但如果黑客们利用似是而非的虚假数据欺骗这些人工智能系统,一定会带来无尽的麻烦。在近期的西班牙巴塞罗那人工智能大会上,OpenAI组织致力于研究欺骗机器学习系统的伊恩·古德菲洛表示,人工智能系统很容易遭受攻击,而且防不胜防。机器学习系统具有辨识数据模式的倾向,利用这一特点可以控制机器学习系统。因为从某种程度上说,机器学习系统没有真
MIT Technology Review Web site December 15, 2016 Published article said that more and more companies are applying advanced artificial intelligence technology, but if hackers use deceptively deceptive data to deceive these artificial intelligence systems, it will certainly bring Endless trouble. At a recent AIIC conference in Barcelona, Spain, Ian Goodfellow, an OpenAI group devoted to studying cheating machine learning systems, said AI systems are vulnerable and unmanageable. Machine learning systems have a tendency to identify patterns of data and use this feature to control machine learning systems. Because to some extent, machine learning system is not true