From Scaling to Epistemic Self-Transformation: Philosophical Limits of Contemporary Artificial Intelligence and Alternative Pathways to Superintelligence

The rapid development of artificial intelligence has renewed fundamental questions concerning the nature of intelligence, the explanatory limits of computational models, and the conditions under which artificial superintelligence might emerge. Contemporary AI research has demonstrated the remarkable effectiveness of large-scale neural networks, computational scaling, optimization, and increasingly sophisticated training procedures. Nevertheless, the success of these approaches does not establish their theoretical completeness or demonstrate that further scaling alone will necessarily produce artificial superintelligence. This paper critically examines the philosophical foundations of contemporary artificial intelligence, focusing on the relationship between reductionism, computationalism, emergence, structural organization, embodied cognition, and open-ended development. It argues that mathematical describability should not be confused with explanatory reductionism, and that intelligence may require analysis across multiple interacting levels of organization. The paper introduces epistemic self-transformation as a conceptual framework for understanding the capacity of an intelligent system to revise not only its solutions and strategies but also its conceptual representations, learning mechanisms, and procedures for evaluating knowledge. Six competing explanatory hypotheses are formulated: the Scaling Sufficiency Hypothesis, Architectural Organization Sufficiency Hypothesis, Enactive–Causal Intelligence Hypothesis, Metaplastic Intelligence Hypothesis, Collective–Relational Intelligence Hypothesis, and Open-Ended Epistemic Evolution Hypothesis. A seventh, cross-cutting hypothesis concerns the necessity of epistemically verifiable self-improvement. The analysis proposes a distinction between intelligence as an existing capacity and intelligence as a capacity-generating developmental process. It suggests that the future of artificial superintelligence research may depend not only on improving computational performance but also on understanding the conditions under which intelligent systems can reliably transform their own methods of acquiring, representing, and validating knowledge. The proposed framework is theoretical rather than empirically established. Its purpose is to clarify competing assumptions, identify unresolved philosophical problems, and provide a foundation for future interdisciplinary investigation.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24426108
Primary Topic
Psychiatry, Mental Health, Neuroscience
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

From Scaling to Epistemic Self-Transformation: Philosophical Limits of Contemporary Artificial Intelligence and Alternative Pathways to Superintelligence

Roman Shaban
Zenodo (CERN European Organization for Nuclear Research)
Psychiatry, Mental Health, Neuroscience
preprint

From Scaling to Epistemic Self-Transformation: Philosophical Limits of Contemporary Artificial Intelligence and Alternative Pathways to Superintelligence

Roman Shaban
preprint en

Abstract

The rapid development of artificial intelligence has renewed fundamental questions concerning the nature of intelligence, the explanatory limits of computational models, and the conditions under which artificial superintelligence might emerge. Contemporary AI research has demonstrated the remarkable effectiveness of large-scale neural networks, computational scaling, optimization, and increasingly sophisticated training procedures. Nevertheless, the success of these approaches does not establish their theoretical completeness or demonstrate that further scaling alone will necessarily produce artificial superintelligence. This paper critically examines the philosophical foundations of contemporary artificial intelligence, focusing on the relationship between reductionism, computationalism, emergence, structural organization, embodied cognition, and open-ended development. It argues that mathematical describability should not be confused with explanatory reductionism, and that intelligence may require analysis across multiple interacting levels of organization. The paper introduces epistemic self-transformation as a conceptual framework for understanding the capacity of an intelligent system to revise not only its solutions and strategies but also its conceptual representations, learning mechanisms, and procedures for evaluating knowledge. Six competing explanatory hypotheses are formulated: the Scaling Sufficiency Hypothesis, Architectural Organization Sufficiency Hypothesis, Enactive–Causal Intelligence Hypothesis, Metaplastic Intelligence Hypothesis, Collective–Relational Intelligence Hypothesis, and Open-Ended Epistemic Evolution Hypothesis. A seventh, cross-cutting hypothesis concerns the necessity of epistemically verifiable self-improvement. The analysis proposes a distinction between intelligence as an existing capacity and intelligence as a capacity-generating developmental process. It suggests that the future of artificial superintelligence research may depend not only on improving computational performance but also on understanding the conditions under which intelligent systems can reliably transform their own methods of acquiring, representing, and validating knowledge. The proposed framework is theoretical rather than empirically established. Its purpose is to clarify competing assumptions, identify unresolved philosophical problems, and provide a foundation for future interdisciplinary investigation.

Zenodo (CERN European Organization for Nuclear Research)
Psychiatry, Mental Health, Neuroscience
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon