The Theory of Verifiable Epistemic Self-Transformation: A Conceptual Model of Artificial Intelligence Development and the Possible Emergence of Superintelligence
This paper proposes a theoretical model of artificial intelligence development based on the principle of Verifiable Epistemic Self-Transformation (VEST). Its central premise is that the development of general intellectual capabilities may depend not only on increasing computational resources, expanding knowledge, or improving the complexity of neural architectures, but also on an intelligent system's capacity to modify its own mechanisms for generating, organizing, evaluating, and applying knowledge.The central hypothesis suggests that the interaction of epistemic plasticity, metaplastic regulation, and independent evaluation of cognitive modifications may support cumulative development of intellectual capabilities in open and changing environments. Unlike approaches primarily concerned with optimizing performance within established architectures or predefined task distributions, the proposed framework examines intelligence as a system capable of transforming the mechanisms underlying its own further development.The paper defines epistemic self-transformation, distinguishes performance adaptation, strategic adaptation, representational restructuring, and metaplastic adaptation, and formulates the fundamental principles of VEST. A functional architecture is proposed, integrating the detection of epistemic discrepancies, restructuring of conceptual representations, active hypothesis testing, knowledge consolidation, and regulation of future learning mechanisms.A preliminary formal representation of the self-transformation process is introduced, together with theoretical predictions and conditions under which the central hypothesis could receive empirical support or be challenged.The paper argues that verifiable epistemic self-transformation may constitute a promising organizing principle for artificial systems capable of sustained development of general intelligence. However, whether this principle is necessary or sufficient for the emergence of artificial superintelligence remains an open question.The proposed framework is theoretical and conceptual. No original empirical experiments are reported.
Authors
- Roman Shaban (ORCID: https://orcid.org/0009-0009-5259-6102)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24426533
- Primary Topic
- Psychiatry, Mental Health, Neuroscience
- Type
- preprint