Substrate Independence vs. Biological Vitalism

Sep 26

Functionalists say mind is what a system does. Biological naturalists say it is what a system is made of. Machine consciousness depends on which side is right, and neither side has a decisive argument.

Every claim that a machine could be conscious rests on an assumption that usually goes unstated: that consciousness depends on organization rather than material. Change the substrate from carbon to silicon, keep the functional organization, and experience comes along. This is substrate independence, and it underwrites nearly all serious work on AI consciousness. The 2023 Butlin and Long report adopted it explicitly — as a working hypothesis, not a finding.

The opposing view is older than computing and has never gone away.

The functionalist case

Functionalism grew out of Hilary Putnam's multiple realizability argument in the 1960s. Pain in a human, an octopus and a hypothetical alien plausibly involve very different physical machinery. If pain were identical to a particular neural state, only creatures with that state could feel it. Since that seems wrong, mental states must be defined by their functional role — what causes them, what they cause, how they relate to other states — rather than their physical makeup.

If that is right, substrate is an implementation detail. A mind is a pattern, and patterns can run on anything capable of instantiating them.

David Chalmers sharpened the argument with his fading qualia thought experiment. Replace a person's neurons one at a time with silicon chips that perform exactly the same input-output function. Behavior never changes; the person keeps reporting vivid experience. If silicon could not support consciousness, experience would have to fade or vanish at some point while the subject insisted, with full functional sincerity, that nothing had changed. Chalmers argues that this systematic disconnect between experience and belief is far less plausible than the alternative: that the functionally identical silicon system is conscious throughout.

Searle's objection

John Searle's biological naturalism rejects the move. Consciousness, he argues, is a biological phenomenon caused by specific neurobiological processes, in the way digestion is caused by stomachs and photosynthesis by chloroplasts. A computer simulation of digestion digests nothing. A simulation of a rainstorm leaves no one wet. Why should a simulation of a brain think?

The core of Searle's position is a claim about computation itself: programs are defined purely syntactically, as symbol manipulation according to formal rules, and syntax is neither constitutive of nor sufficient for semantics. That argument, developed through the Chinese Room in 1980 and restated in the years after, is aimed at the claim that running the right program is enough for a mind. Searle never argued that only biological brains could be conscious. He allowed that an artificial system with the right causal powers could be. His point was that those powers would come from the physics of the machine, not from the program it ran.

Functionalists have pressed Searle on what those causal powers are. He has not specified them, and critics argue that without such a specification the position is an assertion that the brain is special rather than an explanation of why.

The modern version

The strongest contemporary skepticism does not rely on Searle. It comes from predictive processing and the life-mind continuity thesis.

Anil Seth, in Being You (2021), argues that conscious experience is grounded in the brain's predictive regulation of the body — the ongoing, high-stakes work of keeping a living organism within viable bounds. Emotion, selfhood and the felt quality of perception all trace back to that interoceptive modeling. On this account, consciousness may be tied not to biological material as such but to being alive: to metabolism, self-maintenance and a body with something to lose. A language model predicts tokens. It does not predict and regulate its own continued existence, because it has none to regulate.

This is not vitalism in the old sense. Seth does not posit a special life force. He posits that the functions that matter for consciousness may be ones only living systems currently perform — which makes the dispute partly empirical rather than purely philosophical.

Where the debate actually stands

Analysis. Framing this as silicon versus wetware obscures the more useful distinction. Almost no one in the current debate holds that carbon atoms are magic. The real disagreement is over which functions matter and at what grain.

A coarse-grained functionalist says input-output behavior is enough. A fine-grained functionalist says internal causal organization must match too, down to some level of detail. A biological functionalist in Seth's mold says the relevant functions include self-maintenance of a living body. Searle says the relevant level is the physical causal powers of the substrate. IIT, from the previous post, says the same thing in mathematical form.

Each of those positions yields a different verdict on large language models, and none can currently be settled by experiment, because there is no agreed test for the thing being argued about. That is the real problem underneath the substrate debate. Both sides agree consciousness is caused by something physical. They disagree about what, and the only instrument that registers the phenomenon directly is the one each of us has from the inside.

Related reading

→ The Integrated Information Bottleneck

→ The Chinese Room in the Age of Reasoning Models

→ The P-Zombie Dilemma

Social: Nobody serious thinks carbon is magic. The real fight over machine consciousness is about which functions matter — and whether being alive is one of them.

Substrate Independence vs. Biological Vitalism

Global Workspace Theory in Autonomous Systems

The Integrated Information Bottleneck

Your Site Title

email@example.com