Maxwell and Artificial Intelligence: Preliminary QED Models of AI (LLM) Dynamics
Yingrui Yang · Preprints.org · 2025
The present paper aims to set up preliminary conceptual structures of two electrodynamic models of new AI (LLM) dynamics, one is based on the interaction of a charge filed with Maxwell vector potential field, and another is based on the interaction of a spinor field with Maxwell field. For this approach, it introduces a number of new concepts. The AI demand is defined as a two-components unit, a task charge and the corresponding intelligence. Neural network embeds the features of the task into a vector space, which is defined as Maxwell vector potential field. The gradience descent conducts the field strength. For spinor field, it makes the human intelligence and machine intelligence orthogonal; the former is the AI demand and latter the AI supply. The Maxwell field becomes the photon field, which reflects the cost function. The interactions between a pair of AI demand and AI supply and the cost function are pictured by the Feynman Diagram.