人工智能中计算主义的当前难题及转进策略  

The Current Problems and Transition Strategies of Computationalism in Artificial Intelligence

作　　者：黄梦瑶 顿新国[1] HUANG Meng-yao;DUN Xin-guo(Department of Philosophy/Instiute for Contemporary Philosophy of Intelligence&Future of Humanlity,Nanjing University,Nanjing 210023,China)

机构地区：[1]南京大学哲学系/当代智能哲学与人类未来研究中心,南京210023

出　　处：《自然辩证法研究》2023年第11期55-61,共7页Studies in Dialectics of Nature

基　　金：国家社会科学基金重点项目“智能哲学视阈下的归纳逻辑前沿问题研究”(23AZX018);南京大学认知成像揭榜挂帅项目“基于自由能原理的主动推理研究”。

摘　　要：在莱布尼茨、丘奇和图灵等人所奠定的计算的数理理论的基础上,数字计算机诞生,1956年达特茅斯会议将其命名为人工智能。受逻辑与数学的影响,这一时期人工智能主要遵循计算主义的建构方法将认知看作符号计算,但这一做法遭到了强烈的哲学批判,随着日本第五代计算机计划的幻灭,计算主义彻底陷入低谷。事实上,计算主义以其符号表征和符号操纵的方式已然实现了对人的计算智能的建模。由于计算智能是人类智能的重要部分,因此构建强人工智能离不开计算主义。不过这些批判也说明了一个重要问题,即经典计算主义需要新方向、新发展。Building on the mathematical theory of computation laid down by Gottfried W.Leibniz,Alonzo Church,and Alan M.Turing,digital comput-ers were born,which the 1956 Dartmouth Conference named Artificial Intelligence.Influenced by logic and mathematics,AI in this period mainly fol-lowed the constructive approach of computationalism to view cognition as symbolic computation,but this approach was subjected to strong philosophical criticism,and with the disillusionment of Japan's fifth-generation computer project,computationalism completely hit a low point.In fact,computation-alism,with its symbolic representations and symbolic manipulations,has achieved the modelling of human computational intelligence.Since computational intelligence is an important part of human intelligence,computationalism is indispensable for building strong AI.But these critiques also illustrate an im-portant point,namely that classical computationalism needs new directions and developments.

关 键 词：GOFAI 计算智能 人工智能 计算主义 

分 类 号：N031[自然科学总论—科学技术哲学]