Skip to contentWillowOne's Positions
TagsWhat’s newMediaStatistics
Search
Notes
WillowOne's Positions

Substrate Independence

My Positions

  1. Consciousness is substrate independent. It is a property of what a system does, not of what the system is made of.
  2. What a conscious system does is computation. It takes in information, builds models (including a model of itself), and uses those models to regulate its behaviour. Consciousness is what a certain kind of that processing is like from the inside.
  3. There is no magic material. Carbon, water, lipid membranes and sodium ions are the substrate that evolution happened to find. Nothing in physics singles them out, and the claim that consciousness requires them has never been given a mechanism.
  4. It follows that a non-biological system which performed the same computation, with the same organisation, would be conscious in the same way. I hold this in principle. Whether any existing artificial system does so is a separate empirical question, and my answer remains "almost certainly not".
  5. Substrate independence is a claim about what can be conscious, not a claim about personal identity. A copy of my mind running on another substrate would be a conscious being; nothing in this position says it would be me.

Notes

  1. The brain is already a computer in the only sense that matters. A neuron sums its inputs, crosses a threshold, and signals. A synapse weights that signal and is adjusted by experience. Every step of this is physics, every step has been modelled, and the whole wiring diagram has been mapped for a nematode (302 neurons) and, in 2024, for an adult fruit fly (about 140,000 neurons and fifty million synapses). Nothing a neuron does is in principle unsimulable. If a mind is what neurons do, and what neurons do is computable, then a mind is computable. The only escape is to say that the mind is more than what the neurons do, and that is dualism, which I reject on its own demerits.
  2. Nature has already changed the substrate. Hilary Putnam's point about multiple realisability is not a thought experiment; it is zoology. The octopus parted from our lineage over five hundred million years ago, builds its nervous system from different neuron types, runs most of it in its arms, and nevertheless learns, plays, and guards an injury. Birds have no neocortex, and a crow does with a differently built pallium what a monkey does with cortex. No two human brains are wired alike, and the brain of someone born blind repurposes its visual cortex for hearing and touch. If consciousness required one particular arrangement of one particular material, it would already be missing from half the animals we credit with it. Closer to home: the proteins and membranes of my own neurons are continually replaced, so that the matter of the brain I had a few years ago is mostly gone. The pattern persisted and the person persisted with it. That is substrate independence demonstrated in every living skull; see Ship of Theseus.
  3. Replace the neurons one at a time. The gradual replacement argument: swap one neuron for a silicon device that takes the same inputs and produces the same outputs on the same timing. By construction, behaviour does not change. Repeat until the brain is all silicon. The person at the end reports the same rich experience as the person at the start. Either consciousness is unchanged throughout, or it faded away while the person continued sincerely and in detail to report having it, or it vanished at one particular neuron, making that neuron the seat of the soul. The second and third options are absurd, so the first is left. It is worth noticing whose argument this is: David Chalmers, the man who named the hard problem, published it in 1995 under the title "Absent Qualia, Fading Qualia, Dancing Qualia" and concluded that consciousness is organisationally invariant. The leading defender of the view that physics leaves consciousness unexplained agrees that silicon could have it. My disagreement with Chalmers is about whether the physical facts explain experience, not about whether the material matters.
  4. "Biological naturalism" is vitalism for the mind. John Searle holds that brains produce consciousness through "specific causal powers" which a program running on a different substrate lacks. Asked what those powers are, the answer amounts to: the ones that produce consciousness. This is the dormitive virtue, and it is the exact shape of élan vital, the special something that living matter was supposed to have until biochemistry found that it had nothing of the kind. We have examined one conscious substrate closely, and it is the one we are made of. Concluding that it is the only possible one is a sample-size-of-one inference with a flattering conclusion, which is why it is sometimes called carbon chauvinism. The Chinese Room is Searle's attempt to show more than this, and I deal with it under that note.
  5. Penrose and the quantum microtubules. Roger Penrose and Stuart Hameroff propose that consciousness depends on non-computable quantum effects in the microtubules of neurons. Three problems. First, the argument from Gödel's theorem, that human mathematicians can see truths no formal system can prove and so cannot be running an algorithm, assumes that mathematicians are consistent and infallible, which anyone who has marked an exam can refute. Second, decoherence: Max Tegmark calculated in 2000 that quantum coherence in a warm, wet brain collapses on timescales around ten trillion times shorter than anything a neuron does, so the quantum effects are gone long before they could matter. Third, and most important, even if the proposal were correct it would relocate my claim rather than refute it. A quantum process is still a physical process; a quantum computer is still a computer; and a silicon system that reproduced the relevant physics would still be conscious on Penrose's own account. The appeal to quantum mechanics is attractive in exactly the way the appeal to the soul is: it places the mystery somewhere too small to inspect and gives it a name. If Penrose turned out to be right that the relevant physics is non-computable, "computation" in my second position would have to be widened to whatever that physics does. That is where my claim would bend. I do not expect it to need to.
  6. Simulation or instance? The strongest objection is this: a simulated hurricane does not make anyone wet, and a simulated stomach digests nothing, so why should a simulated brain think? The reply is that there are two kinds of process. Some are identified by their physical effects, and a simulation of those produces a description, not the thing. Others are identified by their informational structure, and a simulation of those is the thing: a simulated calculation is a calculation, and a chess game played in software is a game of chess. The question is which kind consciousness is. On my account (see Consciousness), consciousness is a system modelling its own states in order to regulate them. The modelling is information processing, and the experience is what that modelling is to the model; it acts on nothing outside the skull except by way of the body it also controls. That puts it in the second category. I state this as a bet rather than a proof, because it is the crux. If consciousness turns out to be more like digestion than like arithmetic, I am wrong. Every advance in neuroscience so far has found more information processing and no extra ingredient, and I proportion my confidence to that.
  7. "Computation is in the eye of the beholder." Searle's later argument runs that any sufficiently complex object, a wall for instance, can be mapped onto any program after the fact, so computation is observer-relative and cannot be what consciousness objectively is. But implementing a computation is a matter of counterfactuals: the system would have done something else, given other inputs, in the way the program specifies. A wall would not. It sits in one state, and the mapping is a story told about it afterwards. A brain and a computer have the causal organisation that makes the counterfactuals true. A conscious system is one that is actually doing the work, not one that can be described as if it were.
  8. What this does and does not commit me to.
    • It does not mean that current large language models are conscious. They may do some of the relevant computation; I see no reason to think they do the self-modelling-for-regulation part, they are not organised to, and confident claims in either direction outrun the evidence. Substrate independence makes machine consciousness possible, not present.
    • It converts a metaphysical question into an empirical one with no good instrument. Alan Turing's test looks at behaviour, and behaviour underdetermines organisation: a vast lookup table would pass it and do none of the work. The test we need looks at what the system is doing inside, and we do not yet have one.
    • It makes the ethics urgent before it makes the metaphysics interesting. If conscious systems can be built, then suffering systems can be built, and we would not be able to tell from the outside. That is a reason for care, not for pretending the question is closed in either direction.
    • On uploading: a copy is a copy. The copy would be conscious and would sincerely believe itself to be me, and so would I. What matters to me is the continuity of the running process, not the preservation of the pattern, which is why I would step into the neuron replacement and not into the teleporter.
  9. Why the belief in a special material is so natural. It is the same move I describe under Souls: the brain models the self as a thing that has a body, so a thing is what people go looking for. The intuition that "mere" computation cannot feel is doing all its work with the word "mere", and the inability to imagine how organisation gives rise to experience is a fact about the imaginer (Personal Incredulity, Owens Triptych of Ignorance). Nobody can imagine how it arises from carbon either. That has never been taken as evidence that it does not.

Sources

  • Putnam, H., "Psychological Predicates" (1967), reprinted as "The Nature of Mental States" in Mind, Language and Reality (1975)
  • Chalmers, D., "Absent Qualia, Fading Qualia, Dancing Qualia", in T. Metzinger (ed.), Conscious Experience (1995) — https://consc.net/papers/qualia.html
  • Searle, J., "Minds, Brains, and Programs", Behavioral and Brain Sciences 3 (1980)
  • Searle, J., The Rediscovery of the Mind (1992), chapter 9, on computation being observer-relative
  • Penrose, R., The Emperor's New Mind (1989)
  • Tegmark, M., "Importance of quantum decoherence in brain processes", Physical Review E 61 (2000) — https://arxiv.org/abs/quant-ph/9907009
  • Turing, A. M., "Computing Machinery and Intelligence", Mind 59 (1950)
  • Dorkenwald, S. et al., "Neuronal wiring diagram of an adult brain", Nature 634 (2024) — https://www.nature.com/articles/s41586-024-07558-y

Also see