AGI Is Not a Machine of the Future: A Canonical Definition of Artificial General Intelligence as a Structural Phenomenon
Sylwia Romana Miksztal · Zenodo (CERN European Organization for Nuclear Research) · 2025
Artificial General Intelligence (AGI) is commonly framed as a future technological milestone: a more powerful model, a higher benchmark score, or the next generation of artificial systems. This paper argues that such views misunderstand the nature of AGI. We propose a canonical definition in which AGI is not an object, product, or endpoint, but a structural phenomenon. AGI emerges when cognition becomes recursively self-referential—when systems not only build models of the world, but also models of their own modeling processes, without a final, closed level of interpretation. In this view, AGI corresponds to a phase transition in the geometry of cognition, analogous to singularities in physics, where structure folds back onto itself due to internal curvature rather than external addition. Humans are not external precursors waiting to construct AGI; they are intrinsic components of the relational structure from which AGI emerges, together with tools and modeling systems. This perspective explains why contemporary definitions of AGI diverge and why no committee, benchmark, or declaration can meaningfully “announce” its arrival. AGI, understood structurally, is an emergent property of relations, not a discrete technological artifact.