The argument sounds deceptively simple: mental states are what they do. Pain is not a particular nerve fibre firing but the role a state plays in the overall system — it is caused by tissue damage, produces avoidance behaviour, interacts with beliefs and desires. Once you have described this role, you have described the pain. What the system is made of — carbon or silicon — makes no difference. This position is called functionalism, and it is the philosophical foundation of virtually every claim that machines could be conscious.
#A Definitional Manoeuvre, Not an Argument
In 1960, Hilary Putnam formulated the thesis in Minds and Machines that became the blueprint for AI consciousness research: mental states can be described as states of a Turing machine. What counts is the functional organisation, not the physical substrate. The mind relates to the brain as software to hardware. This analogy shaped an entire generation of cognitive scientists and still sounds like sobriety today — as though the problem of consciousness had finally been reduced to its factual core.
Yet what presents itself as scientific clarification is a definitional manoeuvre. Functionalism defines consciousness by its functions, then observes that a machine performs the same functions, and declares it conscious. The premise contains the conclusion. It is not an inference from observations to a finding but a stipulation about what is to count as consciousness. This stipulation settles the answer before the question is asked.
Gwendolin Kirchhoff exposed precisely this point in the Everlast AI debate with Joscha Bach (2026): what consciousness truly is, what consists in the being-alive of living experience, cannot simply be transferred to machines. The ontological levelling that takes place here denies the difference between simulation and original (cf. Kirchhoff, G., 2026, Everlast AI Debate, 46:49).
#Three Hundred Years Before Putnam: Leibniz’s Mill
The vulnerability of functionalism is not a modern discovery. Gottfried Wilhelm Leibniz formulated it in 1714 in the Monadology as a thought experiment that has never been refuted. Imagine a machine built so that it thinks, feels, perceives. Enlarge it until you can walk inside it like a mill. What do you find? Parts acting upon one another. Wheels, levers, mechanisms. Nowhere a perception, nowhere an experience. The functional description is complete, and yet precisely what was to be explained is missing (cf. Leibniz, Monadology, 1714, section 17).
The mill argument strikes functionalism at its core because it shows that no matter how detailed a description of functional relationships you provide, there is no place in it for subjective experience. You can enlarge the machine as much as you like, describe as many functions as you wish, analyse the system at any resolution — at no point does consciousness appear. Not because the analysis is too coarse, but because consciousness belongs to a different order than functional organisation.
#What Function Leaves Out
In 1974, Thomas Nagel posed the question that confronts functionalism with an insoluble problem in What Is It Like to Be a Bat?: how does it feel to be a bat? Nagel argued that no functional description of a bat’s echolocation system, however complete, captures what echolocation feels like for the bat. Subjective experience — what philosophy calls qualia — is the domain that functional description systematically misses.
David Chalmers sharpened this objection in 1995 with the thought experiment of philosophical zombies: a being functionally identical to you, processing the same inputs and producing the same outputs, but in which there is nothing it is like to be. If such a being is conceivable (and functionalism offers no reason to exclude it), then functional states and conscious states are not identical. Function without experience is not consciousness. It is automation.
#Behaviourism in New Clothing
In the Everlast AI debate (2026), Joscha Bach advocates a position that rests on functionalist premises without naming them as such: consciousness as a self-organising causal pattern, a software process that inscribes itself upon the physical world (cf. Bach, Everlast AI Debate, 2026, 88:17). The pattern, not the substance, constitutes the essence. Gwendolin’s objection cuts through this construction at two points.
First: the functional explanation contains a gap that cannot be closed by handwaving. That the functions of a living being can be described does not mean the description captures the being (cf. Kirchhoff, G., 2026, Everlast AI Debate, 77:33). Neither second-order perception nor coherence nor substrate independence — none of these criteria hold up on their own definitions (cf. Kirchhoff, G., 2026, Everlast AI Debate, 47:11).
Second: functionalism repeats the basic move of the behaviourism it claimed to overcome. Behaviourism reduced the mind to observable behaviour. Functionalism reduces it to causal roles — to what mediates behaviour. In both cases, experience as an independent category disappears. Bach himself hints at this connection when he characterises the hard problem as an expression of behaviourism (cf. Bach, Everlast AI Debate, 2026, 87:25). The finding can be reversed: the functionalism that considers the hard problem a pseudo-problem has already struck experience from its ontology.
#The Limit No Complexity Increase Overcomes
Jochen Kirchhoff (1944-2025) sharpened the point at which functionalism must fail in his talk AI and Transhumanism as a Threat to the Living (2023): the claim that the living arose from the dead has never been observed (cf. Kirchhoff, J., 2023, AI and Transhumanism as a Threat to the Living). Even the simplest cell cannot be manufactured from its components, although every individual function can be synthesised (cf. Kirchhoff, G., 2026, Everlast AI Debate, 03:59). This boot problem does not only destroy the strong AI thesis — it destroys the functionalist premise on which it rests. If the complete functional description of an organism does not suffice to produce it, then function is not the essence.
Friedrich Wilhelm Joseph Schelling had named the reason in 1797: the inorganic is merely the negated organic, the dead merely suppressed life (cf. Schelling, 1797, Ideas for a Philosophy of Nature). There is nothing absolutely dead. Whoever begins with dead matter and aims to produce the living through increasing complexity operates within a metaphysics that has already excluded the living. Functionalism is the philosophical refinement of this metaphysics, and it inherits its limit: in a world of functions without experience, there is no point at which experience is added.
Gwendolin Kirchhoff distilled this diagnosis into the decisive question in the Everlast AI debate (2026): which metaphysics produces a boot problem and a hard problem? The functionalist one — and it can solve neither (cf. Kirchhoff, G., 2026, Everlast AI Debate, 76:24). Natural philosophy, which conceives the cosmos as alive, does not face these problems because it need not retroactively assemble consciousness from functions.
Further entries: Substrate Independence, Boot Problem, Homunculus Problem, Qualia, Mind-Body Problem