“有理解力的人工智能”的剖析和AI未来的思考  

Analysis of “AI with Understanding” and Thoughts about the Future of AI

作　　者：沈宏梁 SHEN Hong-liang(Princeton Institute of Intelligent Exploration and Application,Princeton NJ 08540,USA)

机构地区：[1]普林斯顿智能探索和应用研究所,普林斯顿新泽西08540

出　　处：《自然辩证法研究》2021年第7期52-57,共6页Studies in Dialectics of Nature

摘　　要：通过讨论有理解力的人工智能这个概念所存在的问题及其原因,试图说明:人工智能,目前重要的不是停留在哲学概念层面的模糊探讨,而是应该借鉴飞机的设计史,发展自己的基础理论(像飞机的空气动力学);进而,发展一门像空气动力学这样的基础理论,可以借鉴牛顿力学的发展历程,发展更加强大的数学来支撑人工智能的变革。This paper points out that the concept of "AI that understands" is not a good concept, and the most important thing in AI research is not to stay at the discuss on the philosophy level. It should learn from the history of plane design to develop its own aerodynamics. Further more, if AI wants to develop its own aerodynamics, it should learn from Newtonian mechanics and needs more powerful mathematics tool for its revolution.

关 键 词：有理解力的人工智能 不确定性 不可解释 深度学习 新数学 

分 类 号：N031[自然科学总论—科学技术哲学]