Ontological Commitments in Formal Models of Artificial General Intelligence

This paper examines the ontological commitments implicit in the two canonical formal frameworks of Artificial General Intelligence — AIXI and the Gödel Machine — using Quine's criterion of ontological commitment. We propose an ontological signature Ω(F) = ⟨α, ω, μ, ι⟩ characterizing each framework's presuppositions about agency, domain structure, internal states, and the formal criterion of intelligence. The central result is the Dichotomy of Universality (Theorem 4.10): normative universality (D1) is formally incompatible with both formal self-foundation (D2) and computational realizability (D3), partitioning the space of AGI frameworks into two irreconcilable blocks. AIXI satisfies D1 at the cost of incomputability (Turing barrier); the Gödel Machine satisfies D2∧D3 at the cost of incompleteness (Gödel barrier). We argue that the structural limits of each framework are not contingent technical deficiencies but necessary logical consequences of their ontological commitments — a thesis we term ontological determination. The dispute between AGI programs is therefore not merely technical, but grounded in inescapable ontological choices.