人机智能的分界  

Demarcation of human-machine intelligence

作　　者：刘伟[1] LIU Wei(School of Artificial Intelligence,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]北京邮电大学人工智能学院,北京100876

出　　处：《科技导报》2023年第13期118-128,共11页Science & Technology Review

摘　　要：迄今人工智能在智能水平和能力范围上与人类相比仍存在极大的差距,这使得它的发展受到了阻碍。究其原因,人类智能和机器智能之间存在无法跨越的界限。依赖于符号指向对象的机器只能进行封闭环境下的形式化计算,却无法像人那样实现开放环境下的意向性算计,在这种背景下,人机优势互补显得尤为重要。而当前人机的关系主要是硬性的功能分配,未来数字世界中人机关系更可能是柔性的能力分工。探讨了人工智能的瓶颈、智能的第一原理等问题,思考了其对人机问题的影响,指出人机混合智能才是人工智能未来的发展方向的观点。The victory of Deep Blue against Kasparov in 1997 and the victory of Alpha Dog over Lee Sedol in 2016 set off the trend of artificial intelligence,however,artificial intelligence still has so far a huge gap behind humans in terms of intelligence level and range of capabilities,which hinders its development.The reason is that there is an unbridgeable boundary between human intelligence and machine intelligence.In this context,machines that rely on symbols pointing to objects can only perform formal calculations in a closed environment but cannot realize intentional calculations in an open environment like humans.Therefore,the complementary advantages of man and machine are particularly important.The current relationship between man and machine is mainly a hard function allocation,whereas the relationship between man and machine in the future digital world would more likely be a flexible division of capabilities.This paper discusses in detail the bottlenecks of artificial intelligence,the first principle of intelligence,and other issues,and gives more in-depth thinking about human-machine problem amid these problems.It is proposed that human-machine hybrid intelligence is the future development direction of artificial intelligence.

关 键 词：人工智能 功能分配 算计 人机混合智能 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]