统一理论——人工智能研究的新进展

Mechanism Approach:A Unified Theory of AI

人们一直认为结构、功能和行为是智能系统3个基本支柱,因此形成了人工智能理论的结构模拟、功能模拟和行为模拟三大学派。但是,这些研究留下许多问题:互相之间缺乏沟通,却常有"孰优孰劣"的争论。文中提出了一种全新的研究途径——智能生成的"机制方法",希望从智能生成的共性机制入手探讨智能的本质。研究发现:在一般情形下,智能生成机制表现为"信息-知识-智能的转换",由此引出了"知识理论",并发现了知识的生态结构;更为有趣的是,人工智能的结构模拟、功能模拟、行为模拟3种理论在"机制方法"框架内得到了和谐完美的统一。这为智能科学技术的发展提供了新的理解和新的机会。

There have been so far three prestigious schools of AI theory respectively including the structural, functional and behavioral approaches to human intelligence. However, there have been also, controversies among them over the theme of which one of the three should be superior to another. A new approach, mechanism approach, was thus proposed to explore the essence of the intelligence. It is discovered that the mechanism of intelligence formation should be a series of transformations from information to knowledge and further to intelligence. Thus, knowledge theory was introduced and the ecological structure of knowledge was found out. Interestingly, the three existed schools of AI theory can well he unified under the framework of the mechanism approach. The mechanism approach and the unification of AI may provide a new foundation for the future development of AI theory and applications.