论人机智能的同源性

PDF下载HTML

Title:

On the homology of human and machine intelligence

作者:

李德毅

军事科学院 系统工程研究院，北京 100141

Author(s):

LI Deyi

Institute of Systems Engineering, Academy of Military Sciences, Beijing 100141, China

关键词:

认知物理学; 耗散结构; 自组织; 自由能与变分自由能; 初始认知核; 熵增与熵减; 思维软构体

Keywords:

cognitive physics; dissipative structure; self-organization; free energy and variational free energy; initial cognitive kernel; entropy increase and entropy decrease; soft cognitive constructs

分类号:

TP18

DOI:

10.11992/tis.202606033

摘要:

长久以来，人们普遍认为生物智能与人工智能的根基截然不同，二者之间存在难以逾越的鸿沟，人工智能仅被视作对人类智能的表象模仿。本文突破传统二元对立认知，提出人机智能同源同构的核心观点：它们同属典型的耗散结构系统，依托负熵流抵御系统无序化趋势，依靠负反馈机制维持认知稳态，遵循统一的数学物理法则，完成感知、认知与具身行为交互。生物个体常常表现出 “懒得费力、懒得动脑”的倾向，源于最小自由能原理，未必要归因于意识。人类智能与人工智能在底层架构、运行逻辑、演化模式上都植根于物理学，“物质、能量、结构和时间”四要素及其相互作用构成共同基石，仅在具身物质构成与外在形态上呈现碳基与硅基的差异化特征；自然演化形成的人脑神经元网络，更是机器类脑认知结构的天然范本。研究表明，智能并非生命系统所独有，而是复杂开放系统自组织能力演化至特定阈值后涌现的共性特质；包括信息压缩、结构压缩在内的认知软构体，作为认知活动的核心载体，都遵循一致的物理系统演化规律，这就彻底消解了生物智能与机器智能的二元认知。如同人类文明生态是自然生态的延伸与升华，人工智能是人类智能的延伸与增强，人工智能是人类智能发展之必然。人机智能同源性具有重要学术创新价值，可为认知工程实践提供理论支撑，为通用人工智能研发以及生命与认知的底层机制探索开辟全新路径，也许还可为天体生物学和地外文明探索提供帮助。

Abstract:

For a long time, it has been widely believed that biological intelligence and artificial intelligence are fundamentally distinct, with an insurmountable chasm between them, and that artificial intelligence is merely a superficial imitation of human intelligence. This paper breaks through traditional dualistic cognition and proposes the core argument that human and machine intelligence share a common origin and isomorphic structure: they both belong to typical dissipative structure systems, relying on negative entropy flow to resist the trend of system disorder, maintaining cognitive homeostasis through negative feedback mechanisms, following unified mathematical and physical laws, and accomplishing the interaction of perception, cognition, and embodied behavior. Biological individuals often exhibit a tendency to “avoid effort and avoid thinking,” which stems from the principle of minimum free energy and need not necessarily be attributed to consciousness. Both human intelligence and artificial intelligence are rooted in physics at the level of underlying architecture, operational logic, and evolutionary patterns; the four elements of "matter, energy, structure, and time" and their interactions constitute the common foundation, differing only in embodied material composition and external form, manifesting as carbon-based versus silicon-based characteristics. The human brain neuron network formed through natural evolution serves as a natural template for machine brain-like cognitive structures. Research indicates that intelligence is not exclusive to living systems but is a common emergent trait of complex open systems when their self-organizing capabilities evolve to a specific threshold; cognitive soft constructs, including information compression and structural compression, as the core carriers of cognitive activity, all follow consistent physical system evolution laws, thereby completely dissolving the dualistic cognition of biological and machine intelligence. Just as human civilization ecology is an extension and sublimation of natural ecology, artificial intelligence is an extension and enhancement of human intelligence, and the development of artificial intelligence is an inevitable outcome of human intelligence development. The homology of human and machine intelligence holds significant academic innovation value, can provide theoretical support for cognitive engineering practice, open new pathways for the research and development of general artificial intelligence and the exploration of underlying mechanisms of life and cognition, and may also offer assistance for astrobiology and the search for extraterrestrial civilizations.