人工智能研究进路及其整合的通用智能机制启示  

General Intelligence Mechanism Revelations of Research Approaches and Its Integration of Artificial Intelligence

作　　者：王天恩[1,2] WANG Tian’en

机构地区：[1]上海大学智能哲学与文化研究院,上海200444 [2]上海大学哲学系

出　　处：《北华大学学报（社会科学版）》2024年第4期35-44,151,共11页Journal of Beihua University(Social Sciences)

基　　金：教育部哲学社会科学研究重大课题攻关项目“人工智能的哲学思考研究”(18JZD013)的阶段性研究成果。

摘　　要：迄今为止,就类人智能意义上的通用人工智能核心机制而言,还没有形成真正可行的研究进路,但从信息层次可以清楚地看到,现有人工智能研究进路及其整合却都有程度不同的通用智能机制启示。符号主义进路主要涉及通用人工智能的机器机制基础,其启示主要是人类知识的符号表征和运演等在通用人工智能中不可或缺的地位;联结主义进路的启示主要在于智能体内部关系在人工智能中的地位,涉及基于内部结构的解释机制基础的重要性和局限;行为主义进路的启示主要涉及诠释及其与解释之间的具体关系机制;而人工智能研究进路整合的启示则是解释和诠释构成的双向循环理解。As of now,a truly viable research approach for the core mechanism of artificial general intelligence in the sense of humanoid intelligence has not been established.However,it is clear from the information level,existing research approaches in artificial intelligence and their integration offer varying degrees of revelations into general intelligence mechanisms.The symbolic approach primarily involves the foundational machine mechanisms of artificial general intelligence,with revelations focusing on the indispensable role of symbolic representation and manipulation of human knowledge in artificial general intelligence.The connectionist approach provides crucial insights into the internal relations within intelligent agents in artificial intelligence,addressing the importance and limitations of the basis of explanative mechanisms and the gaps or limitations in their mechanistic foundations and interpretations based on internal structures.The behaviorist approach mainly revolve around interpretation and the specific relation mechanisms between it and explanation.The significant insight into integrated general intelligence mechanisms in artificial intelligence research approaches is the bidirectional cyclical understanding formed by the interplay of interpretation and explanation.

关 键 词：人工智能 信息 通用智能机制 符号主义 联结主义 行为主义 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]