第15卷第3期 智能系统学报 Vol.15 No.3 2020年5月 CAAI Transactions on Intelligent Systems May 2020 D0L:10.11992tis.202002019 “范式变革”引领与“信息转换”担纲：机制主义 通用人工智能的理论精髓 钟义信 (北京邮电大学人工智能学院，北京100876) 摘要：世界人工智能研究至今一直沿用着物质科学的科学范式（科学观和方法论），因此被分解为结构主义人 工智能（人工神经网络）、功能主义人工智能（物理符号系统/专家系统）和行为主义人工智能（感知动作系统/智 能机器人)3个各自为战互不相容的学派。虽然各个学派都获得了一些精彩的局部性专用性成果，却没有通用 性整体性的人工智能应用，更无法形成通用的人工智能整体理论，这成为人工智能研究与发展的最大痛点。目 前，通用性整体性的人工智能理论越来越成为社会的紧迫需求。为此，本文依据作者四十多年研究的积累.总 结和提出了“机制主义通用人工智能理论”，特别强调了“范式变革”和“信息转换”，希望引起学界的研讨和批评。 关键词：人工智能；范式变革；信息转换；整体理论；通用理论；智能生成机制；显因素；隐因素 中图分类号：TP18文献标志码：A文章编号：1673-4785(2020)03-0615-08 中文引用格式：钟义信.“范式变革”引领与“信息转换”担纲：机制主义通用人工智能的理论精髓，智能系统学报，2020， 15(3):615-622. 英文引用格式：ZHONG Yixin.Leading of paradigm shift and undertaking of information conversion::theoretical essence of mech- anism-based general Al[JI.CAAI transactions on intelligent systems,2020,15(3):615-622. Leading of paradigm shift and undertaking of information conversion: theoretical essence of mechanism-based general Al ZHONG Yixin (AI School,Beijing University of Posts and Telecommunications,Beijing 100876,China) Abstract:Because of the employment of scientific paradigm (i.e.,scientific outlook and methodology)in the field of materials science,the global research on artificial intelligence(Al)has been divided into three branches,namely,struc- turalism-based AI(e.g.,artificial neural networks),functionalism-based AI (e.g.,physical symbol systems and expert systems),and behaviorism-based AI(e.g.,sensorimotor systems and intelligent robots),which are independent from each other and mutually unharmonious.A decent number of results from each of the three branches have been achieved; however,no progress has been made in the global theory of Al,let alone the universal theory of Al,and this has be- come the biggest pain point in the research and development of Al.Presently,the universal and global theories of AI have gradually become urgent social demands.Accordingly,the article titled "The theory of mechanistic general artifi- cial intelligence"has been presented on the basis of the research experiences of the author during the past four decades; the work mainly focuses on"paradigm shift"and "information conversion."The author hopes that the views and results presented in the article could draw discussions and criticisms from the readers of the AI academic circle. Keywords:artificial intelligence(AI),paradigm shift,information conversion;global theory;universal theory;mechan- ism for intelligence growth;explicit factor,implicit factor 收稿日期：2020-02-26. 科学范式引发的思考。在理想的情况下，为 基金项目：国家社科基金项目(I8ZDA027):国家自然科学基金 了对某个学科展开深入系统的科学研究，首先需 项目(60873001.60575034). 通信作者：钟义信.E-mail:zyx@bupt.edu.cn. 要对这个学科有一个整体性的理解，能够在宏观DOI: 10.11992/tis.202002019 “范式变革”引领与“信息转换”担纲：机制主义 通用人工智能的理论精髓 钟义信 （北京邮电大学 人工智能学院，北京 100876） 摘 要：世界人工智能研究至今一直沿用着物质科学的科学范式 (科学观和方法论)，因此被分解为结构主义人 工智能 (人工神经网络)、功能主义人工智能 (物理符号系统/专家系统) 和行为主义人工智能 (感知动作系统/智 能机器人) 3 个各自为战互不相容的学派。虽然各个学派都获得了一些精彩的局部性专用性成果，却没有通用 性整体性的人工智能应用，更无法形成通用的人工智能整体理论，这成为人工智能研究与发展的最大痛点。目 前，通用性整体性的人工智能理论越来越成为社会的紧迫需求。为此，本文依据作者四十多年研究的积累，总 结和提出了“机制主义通用人工智能理论”，特别强调了“范式变革”和“信息转换”，希望引起学界的研讨和批评。 关键词：人工智能；范式变革；信息转换；整体理论；通用理论；智能生成机制；显因素；隐因素 中图分类号：TP18 文献标志码：A 文章编号：1673−4785(2020)03−0615−08 中文引用格式：钟义信. “范式变革”引领与“信息转换”担纲：机制主义通用人工智能的理论精髓 [J]. 智能系统学报, 2020, 15(3): 615–622. 英文引用格式：ZHONG Yixin. Leading of paradigm shift and undertaking of information conversion: theoretical essence of mech￾anism-based general AI[J]. CAAI transactions on intelligent systems, 2020, 15(3): 615–622. Leading of paradigm shift and undertaking of information conversion: theoretical essence of mechanism-based general AI ZHONG Yixin (AI School, Beijing University of Posts and Telecommunications, Beijing 100876, China) Abstract: Because of the employment of scientific paradigm (i.e., scientific outlook and methodology) in the field of materials science, the global research on artificial intelligence (AI) has been divided into three branches, namely, struc￾turalism-based AI (e.g., artificial neural networks), functionalism-based AI (e.g., physical symbol systems and expert systems), and behaviorism-based AI (e.g., sensorimotor systems and intelligent robots), which are independent from each other and mutually unharmonious. A decent number of results from each of the three branches have been achieved; however, no progress has been made in the global theory of AI, let alone the universal theory of AI, and this has be￾come the biggest pain point in the research and development of AI. Presently, the universal and global theories of AI have gradually become urgent social demands. Accordingly, the article titled “The theory of mechanistic general artifi￾cial intelligence” has been presented on the basis of the research experiences of the author during the past four decades; the work mainly focuses on “paradigm shift” and “information conversion.” The author hopes that the views and results presented in the article could draw discussions and criticisms from the readers of the AI academic circle. Keywords: artificial intelligence (AI); paradigm shift; information conversion; global theory; universal theory; mechan￾ism for intelligence growth; explicit factor; implicit factor 科学范式引发的思考。在理想的情况下，为 了对某个学科展开深入系统的科学研究，首先需 要对这个学科有一个整体性的理解，能够在宏观 收稿日期：2020−02−26. 基金项目：国家社科基金项目 (18ZDA027)；国家自然科学基金 项目 (60873001，60575034). 通信作者：钟义信. E-mail：zyx@bupt.edu.cn. 第 15 卷第 3 期 智 能 系 统 学 报 Vol.15 No.3 2020 年 5 月 CAAI Transactions on Intelligent Systems May 2020