A Causal-Structure Theory of the Micro-to-Macro Transition and an AI-Based Method for Discovering New Physical Laws

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

This work proposes a new micro-to-macro theory of how deterministic microscopic causal structures give rise to macroscopic objects, macroscopic memory, temporal directionality, and macroscopic laws. Unlike approaches that begin with statistical averaging, coarse-graining, or predefined macroscopic variables, the theory takes microscopic causal structures themselves as the primary objects of analysis. It focuses on causal organizations that can re-enter the formation of structures of the same class and accumulate through system evolution, thereby providing a route from microscopic causal organization to stable macroscopic carriers and macroscopic laws. On the basis of this theory, the work further develops a method for AI-assisted discovery of new laws. The theoretical principles are translated into an executable research workflow in which an AI system starts from a concrete microscopic model, identifies causally accumulative structures, constructs candidate macroscopic carriers, detects causal structures shared by different carriers, and systematically derives candidate macroscopic relations, falsifiable predictions, and potentially new laws. These candidates can then be subjected to literature checks, counterexample analysis, numerical investigation, and experimental validation. The aim of this work is therefore twofold: first, to introduce a new theoretical framework for deriving macroscopic organization and laws from microscopic causal structure; and second, to transform that framework into a repeatable, scalable, and AI-executable methodology for exploring previously unknown macroscopic laws. In this sense, AI is used not only to analyze existing physical laws, but to search for new ones under an explicit theoretical framework.As a preliminary result, using this prompt and workflow, ChatGPT under the Sol high-intensity Work configuration has generated and derived nearly one hundred physically meaningful candidate laws. For a substantial fraction of these candidates, preliminary checks have not identified clear counterparts in the existing literature, suggesting that some may represent previously unreported physical laws. This preliminary result further suggests that the framework may enable AI to systematically explore regions of physical law space that have not yet been reached by conventional research workflows. By T

Links and sources

Comments