Intelligence extends itself by creating new meanings and purposes
Co-Recursive Semantic Self-Extension: A Mathematical Foundation for Open-Ended Intelligence
Artificial Intelligence
Summary
Intelligence usually learns and solves problems within fixed ideas and goals, which limits how much it can grow or adapt. The authors explain that true open-ended intelligence needs to go beyond just working with existing ideas; it should be able to create new meanings, questions, and goals by exploring what can't be fully decided or solved yet. They use mathematical and philosophical ideas to describe this process and show why fixed systems can’t fully generate new complexity on their own. They suggest that intelligence grows by continually expanding its own framework instead of staying within old limits.
What this means in practice
- •For machine learning engineers: Design models that can dynamically expand their representation space to create new concepts beyond fixed training domains.
- •For distributed systems developers: Develop systems combining multiple agents with persistent shared memory to increase collective generative capacity beyond individual fixed limits.
A theory result. No direct application yet.
Authors
Simon Hegelich, Bardh Prenkaj, Habiba Sarhan, Morteza Shahrezaye, Kolja Hegelich, Eric Cambria, Christian Montag
Abstract
Artificial intelligence has advanced through scale, data, and computation, yet remains bounded by its representational framework. We ask what makes intelligence open-ended. Every inferential system selects meaningful distinctions, limiting what it can learn. Within a fixed purpose, performance reaches a ceiling set by its semantic domain. Teleological succession, in which purposes arise from earlier ones, faces a similar limit: Boolean-closed purpose-transformations exhaust their reachable semantic territory. We shift from purposes to inferential systems. Drawing on Kant and Gödel, we model cognitive states as partial evaluations over a formal language. Unresolved propositions may be undecided or undecidable. Reasoning can resolve the former but not the latter, leaving a frontier for extension. Non-trivial reductions of undecidable propositions expose undecidable intermediate premises whose reflective reification supports multiple decisive extensions. Finite trace-guided generation remains fallible, requiring meta-level selection and polycontextural exploration. Using Kolmogorov complexity, we show that no architecture with a fixed state description, transition mechanism, and initial condition can generate histories incorporating irreducible algorithmic information. Because physical systems are fixed in this sense, we examine collectives. Coupling finitely many finite agents does not remove the bound. Extension requires persistent external memory preserving semantic content and historical coordinates. Collective generative capacity depends on balancing overlap for transfer with difference for generation. We propose semantic self-extension as a design principle: intelligence should not merely resolve ambiguity within existing meanings, but transform unresolved structures into new spaces of meaning, generating new distinctions, questions, representations, and purposes.