From Physical Observer to Agent: A First-Principles Derivation of the Minimal Complete Architecture

This paper proposes a unified formal framework establishing the structural isomorphism between the "observer" in physics and the "agent" in artificial intelligence. Grounded in cross-disciplinary first principles, we advance the core proposition that an agent is fundamentally an open information-processing system, and rigorously derive its Minimal Complete Architecture (MCA) through the state-space coverage method. This architecture comprises five functionally irreducible and logically necessary dimensions: Input (I), Output (O), Memory (M), Generation (G), and Control (C). As a preliminary theoretical exploration, the proposed paradigm demonstrates significant potential for breakthrough contributions across three fronts: first, it provides a substrate-independent analytical framework spanning physical, living, and artificial systems; second, it furnishes operationally tractable and experimentally falsifiable foundational definitions of intelligence and consciousness; third, it offers an original theoretical perspective explaining the divergences among classical mechanics, relativity, and quantum mechanics through the lens of varying observer capability configurations. Serving as the theoretical point of departure for this emerging paradigm, the framework delineates a clear agenda for future research to provide increasingly rigorous and complete answers to these foundational scientific questions.