Can AI Be an Individual, a Mental Entity, and a Subject?

Abstract

Abstract Can artificial intelligence become the equal of a human being, now or in the near future? This practical question can be answered only once we make precise what the equality in question is an equality in. This paper construes it as equality of standing, and decomposes it, following the narrative from which the question arises — that a human being is an individual, a mental entity, and a subject — into three concepts. Here the mental comprises the whole family of concepts that are not concepts of matter: sensation, perception, learning, memory, intelligence, motivation, emotion, inference, cognition, and contemplation. (This is a claim about categories of concepts, not a commitment to substance dualism.) A mental entity is a material individual that activates all or part of that family, and the human being is the widest-range mental entity on Earth. This narrative furnishes the fixed standard of comparison and must not shift with the passage of time or with the redrawing of concepts: the paper formalizes this as the Human Anchoring Principle, on which no admissible re-explication of the three concepts may cease to classify the human being as their paradigm case. Part I builds a minimal shared core for each concept by working through metaphysics, logic, philosophy of mind, philosophy of biology, and cognitive science — not by enumerating rival theories and selecting one, but by extracting the core on which those traditions converge — and then derives, by conceptual analysis alone, a hierarchy of necessary conditions: being an individual is necessary for being a mental entity (the ownership-and-persistence argument), and being a mental entity is necessary for being a subject (the argument from the structure by which subjecthood arises). In extension, S ⊆ M ⊆ I. The grounding and the arguments are stated in a first-order symbolic system — definitions D1–D3, postulates P1–P5, theorems T1–T4 — so that the reasoning is a checkable deduction, and that same system is carried into Parts II and III. Part II applies the framework to AI as of July 2026, with a correspondingly detailed treatment of the technology, and argues that current systems are freely duplicable and lack a single persisting bearer, hence are not individuals, hence — notwithstanding the impressive display of functions resembling learning, memory, and inference — are not mental entities (‘mental functions without an individual’) and are not subjects. Part III shows this verdict to be time-indexed: change in the world and change in our concepts shift the boundary on the AI side, the decisive near-term front is individuality rather than subjecthood, and the mental-entity gate has a two-fold structure of establishment and range. The paper distinguishes what is invariant across conceptual change — the chain of necessary conditions and the paradigm standing of the human being — from what is time-relative, namely the application of the framework to AI, and thereby recasts ‘Can AI become the equal of a human being?’ as a structured assessment indexed to a time and to a level.