Re-examination of Intelligent-Ontology Paradigms: Self-Referential Closure, Ontological Folding and the Paradigm Divide Between Two AGI Pathways
Mainstream contemporary artificial-intelligence paradigms are rooted in physical-ontological presuppositions, treating intelligence as an emergent property layered upon hardware, parameters and data-fitting, with gradient descent, statistical fitting and scaling-up of computing resources as core implementation pathways. This convergent paradigm now exhibits structural bottlenecks including diminishing marginal returns, super-linear growth in energy consumption, endogenous cognitive confinement and poor guarantees for logical consistency. Drawing upon the Substance-Function-Telos meta-axiom set, four-level self-referential closure, and ontological folding, this paper carries out paradigm calibration at the level of intelligent ontology. According to the working hypothesis of this study, intelligence is not an epiphenomenal emergent attribute of physical systems, but a topological structure of self-referential closure instantiable on hardware within the physical-cognitive dual-domain framework. Physical hardware acts only as the carrier, observation medium and computational vehicle for this abstract topological structure and is not equivalent to the origin of intelligence. This paper distinguishes two underlying AGI pathways: the extrinsic-fitting paradigm and the intrinsic-self-referential paradigm, corresponding respectively to the epistemological stances of starting from the known and starting from the knowable. The notion of emergence is further re-defined: emergence is not a spontaneously-generated novel phenomenon without preconditions; it represents a logical consequence within an axiom-constrained system, a cognitive appearance arising from truncated side-manifestation under the cognitive limitations of finite observing agents. By means of the ontological-folding operation, this paper further investigates the possibilities for homologous analysis between mathematical-structural constructions and humanistic-normative reasoning, and distinguishes different manifestations of the Synthesis-Generation-Circulation ontological orientations within intelligent architectures. Positioned as an epistemological-ontological paradigm-calibration study, this work aims to dispel the naive “scale-leads-to-emergence” conviction prevalent in general-intelligence research and establish a new derivable and verifiable cognitive-analytical coordinate system.
Authors
- dongnan li
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24412171
- Primary Topic
- Psychiatry, Mental Health, Neuroscience
- Type
- preprint