机制主义人工智能理论——一种通用的人工智能理论  

Mechanism-based artificial intelligence theory: a universal theory of artifical intelligence

作　　者：钟义信[1] 

机构地区：[1]北京邮电大学计算机学院,北京100876

出　　处：《智能系统学报》2018年第1期2-18,共17页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金项目(60873001;60575034;60496327;69982001;69171023;68872014)

摘　　要：现行人工智能研究取得了许多进展,但存在"深度上浅层化、广度上碎片化和体系上封闭化"的重要缺陷。这不是改进算法或者提高硬件性能所能解决的问题,而是要在科学观方法论上寻找根源。本文依据"科学观→方法论→研究模型→研究途径→基本概念→基本原理"这个顶天立地的研究纲领,总结了信息科学的科学观,提炼了信息生态方法论;在新的科学观和方法论指导下构筑了体现智能生长全过程的研究模型,发现了智能生长的共性机制,确立了机制主义研究途径,进而澄清和匡正了信息(特别是语义信息)、感知、知识、认知、基础意识、情感、理智、综合决策等一系列基础概念,总结了实现信息-知识-智能转换的一组基本原理,创建了机制主义人工智能理论。而且证明了:长期三分而立的结构主义(人工神经网络)、功能主义(专家系统)、行为主义(感知动作系统)三大人工智能理论可在机制主义人工智能理论框架内实现和谐统一;机制主义是生成基础意识、情感、理智三位一体高等人工智能的科学途径;机制主义人工智能理论是通用型的人工智能理论。While artificial intelligence（AI） research has made a great deal of progress, this field also faces severe drawbacks, including being shallow in depth, piece-like in width, and isolation among the pieces. These drawbacks cannot be overcome by optimizing algorithm design and improving hardware performance. The real causes of AI＇s serious problems arise from its scientific view and improperly employed methodology. Following the general guidelines for scientific research architecture, i.e., ＂scientific view and methodology on the top, concepts and principles at the bottom, and the model and approach in between, ＂ in this paper, we summarize the scientific view of AI and define the methodology of information ecology. In accordance with the methodology above, information research cannot stop at the information level, but must continue to the product levels-the knowledge and intelligence levels. Based on this understanding, we design a subject-object interaction model for AI research, propose a mechanism approach to AI, identify and re-build the concepts and principles underlying the entirety of information science, and establish a mechanism-based AI theory. Interestingly, we found that the three AI schools of thought that have operated in isolation from each other for decadesstructuralism-based AI（the artificial neural network）, functionalism-based AI（the expert system）, and behaviorismbased AI（the sensor-motor system）-become harmonious components within the mechanism-based AI theory. Moreover,we also prove that the mechanism-based AI approach is also appropriate for the advanced AI work that unifies the primary consciousness, emotion, and intellect. We thus conclude that the mechanism-based AI theory provides a general theory of artificial intelligence.

关 键 词：信息生态方法论 智能生长机制 机制主义研究途径 通用型人工智能理论 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]