Theoretical Foundations of Artificial General Intelligence Based on the Logic of Life Evolution

Current research on artificial general intelligence (AGI) is deeply constrained by the data-driven paradigm, whose fundamental predicament lies in its forgetting of the biological foundations and evolutionary dimension of intelligence. By tracing the evolutionary history of consciousness from single-celled life to the human spiritual world, this paper argues that the essence of consciousness is a “self/non-self” boundary perception system opened up by living organisms to maintain their own homeostasis, and reveals the operating mechanism of the central nervous system as a predictive coding engine as well as its “bijective prison” predicament [1–3]. On this basis, this paper points out that the birth of language is not merely a revolution in communication, but rather, through a mechanism of symbolic disembedding, enables the mind to break free from the rigid chain of stimulus–response and to ascend from “presence”-bound consciousness to a reflexive spiritual world [4–6]. Accordingly, this paper offers a systematic critique of the “floating symbol” architecture of current large language models and proposes a three-layer nested theoretical framework for AGI—an ontological security layer, an embodied prediction layer, and a symbolic deliberation layer—arguing that the construction of AGI must follow the developmental order of “embodiment first, language second,” with “constitutional monarchy” as the adjudicative principle for inter-layer conflicts. This paper attempts to provide AGI with an ontological foundation rooted in the philosophy of life and an operable architectural blueprint.