人工认知的语境建构与适应性表征解释  

Contextual Constructs and Adaptive Representational Interpretations of Artificial Cognition

作　　者：魏屹东[1] WEI Yidong(School of Philosophy and Sociology,Shanxi University,Taiyuan 030006,China)

机构地区：[1]山西大学哲学社会学学院,山西太原030006

出　　处：《山东科技大学学报（社会科学版）》2023年第1期1-17,共17页Journal of Shandong University of Science and Technology(Social Sciences)

基　　金：国家社会科学基金重大项目“人工认知对自然认知挑战的哲学研究”(21&ZD061);国家社会科学基金后期资助重点项目“人工智能:从物理符号操作到适应性表征”(21FZXA006)。

摘　　要：认知科学的发展使自然认知得到充分研究,人工智能的发展使人工认知成为广泛关注的焦点。然而,这两种认知形式是何种关系,如何关联,如何进行推理,科学上存在着“探索黑箱”,哲学上存在着“解释鸿沟”。鉴于认知发生机制的复杂性和解释上的困难,这里将适应性表征作为认知发生的内在机制和解释框架,据此来探讨人工认知与自然认知的统一认知架构、认知推理的形式表征、概念与语境模型。研究试图表明,认知系统,无论是自然的,还是人工的,均是适应性表征系统,具有自适应、自复制、自组织的自主性和能动性;认知主体,无论是碳基生物,还是硅基装置,都是适应性实体,其行为都是适应性认知。因此,认知或智能是适应性和表征性的统一,适应性表征可合理地说明或解决认知或智能生成问题。这意味着,在适应性表征框架下,人工认知通过模拟自然认知的结构、功能和行为,生成像人类认知具有的通用智能,这可能是新一代人工智能发展的方向。具体而言,这里使用基于语境的适应性表征方法论来修正通用认知架构与认知推理模型,将人工主体分为逻辑主体、搜索主体、决策主体、学习主体和问题-解决主体,建构了一个形式化的语境模型,力图解释或解决科学和哲学普遍关注的智能生成问题,并探讨了其中蕴涵的本体论、认识论和方法论意义。The development of cognitive science has led to the full study of natural cognition, and the development of artificial intelligence has made artificial cognition the focus of extensive attention. However, there is a “black-box of inquiry” in science and an “explanatory gap” in philosophy as to how these two forms of cognition are related and how they are reasoned about. In view of the complexity of the cognitive mechanism and the difficulty of explanation, adaptive representations are used here as an intrinsic mechanism and an explanatory framework for cognition to explore the unified cognitive architecture of artificial and natural cognition, formal representations of cognitive reasoning, and conceptual and contextual models. It is shown that cognitive systems, whether natural or artificial, are adaptive representational systems with autonomy and dynamism which are self-adapting, self-replicating and self-organizing;cognitive agents, whether carbon-based organisms or silicon-based devices, are adaptive entities whose behaviors are adaptive cognition. Thus, cognition or intelligence is a unification of adaptation and representation, and adaptive representations can reasonably account for or solve cognitive or intelligence-generating problems. This means that in the framework of adaptive representation, artificial cognition generates general intelligence like human cognition by simulating the structure, function, and behavior of natural cognition, which may be the direction of the development of next-generation artificial intelligence. Specifically, a context-based adaptive representation methodology is used to revise the general cognitive architecture and cognitive reasoning model;the artificial agents are divided into logical agent, searching agent, decision-making agent, learning agent, and problem-solving agent;a formal contextual model is constructed to explain or solve the intelligence generating problems in science and philosophy. Meanwhile, the ontological, epistemological, and methodological impli

关 键 词：认知系统 认知架构 逻辑建构 语境模型 适应性表征 人工智能 

分 类 号：B81[哲学宗教—逻辑学] TP391[自动化与计算机技术—计算机应用技术]