Current AGI research remains mired in the bottleneck of “digital centrism” characterized by parameter fitting, failing to explain how microscopic units emerge into macroscopic intelligence. The core deficiency lies in the absence of a middle-layer ontology bridging these two levels. Drawing upon the first principles of complex systems, this paper proposes a General Agent Dynamics theory. Its pivotal breakthrough lies in ontologizing the intelligent middle layer into “Domains,” establishing them as independent causal agents that bridge microscopic foundations and macroscopic behaviors. The theory elucidates the mechanisms of generation, steady-state maintenance, and evolution of these entities driven by “operational conflict minimization,” and constructs a three-layer nested architecture—“Core-Module-Domain”—centered on domains. This research bridges the paradigmatic gap from “numerical fitting” to “structural evolution,” filling the “ontological void” in intelligence science and providing a novel dynamical foundation for AGI architecture design, behavioral interpretation, and system collaboration. 现有AGI研究深陷“数字中心主义”的参数拟合瓶颈,将智能还原为可数值化、可参数拟合的静态优化问题,忽视智能作为复杂系统的动态结构演化本质,无法解释微观单元如何涌现为宏观智能,核心症结在于缺失了连接二者的中间层本体。本文基于复杂系统第一性原理,提出通用智能体动力学理论,核心突破在于将智能中间层本体化为“领域”,确立了其作为连接微观基础与宏观行为的独立因果主体地位。 理论阐明了该实体在运行冲突最小化驱动下的生成、稳态维持与演化机制,构建了以领域为核心的“内核-模组-领域”三层嵌套架构。该研究跨越了从“数值拟合”到“结构演化”的范式鸿沟,填补了智能科学的“本体缺口”,为AGI架构设计、行为解释与系统协同提供了全新的动力学基础。
Zhili Ling (Tue,) studied this question.