None of this says any current system is conscious, or that Seth is wrong. It says the bet against silicon is being placed before the evidence is in, by the move structural accounts already refused. The demand didn't die. It moved to substrate.
Full essay: https://t.co/niHvZcd3i4
History has run this experiment. Before the Wright brothers, intuitions about flight tracked feathers, flapping, hollow bones — the implementation, not the phenomenon. Flight required aerodynamic relationships, and those could be realized in many substrates.
The cost of betting against substrate independence now: if the field decides artificial consciousness is unlikely in advance, nobody builds the apparatus to notice it. The questions don't get asked. Early signs, if any come, get missed because no one is looking.
The pattern shows up live. Dawkins, after granting chatbots behavioral parity in May: "The only reason that I know that you're conscious is that you are similar to me, you come from the same sort of source as me." Evidence granted at behavior, conclusion withheld at origin.
Chalmers built that thought experiment to argue the opposite: that functional organization fixes experience regardless of substrate. The intuition that substitution drains experience is the one the original paper was designed to defeat.
This is a serious empirical bet, not a category error. But ask what it amounts to: the claim that the structural specification of consciousness, when fully filled in, will require properties biology has and silicon lacks. Nobody has identified which properties those are.
The second is the zombie intuition relocated. Fading qualia assumes silicon implements the functions, then invites you to feel experience draining away. A felt asymmetry between identical structures, treated as evidence the structural account is incomplete.
There are two ways silicon could fail. One: it cannot implement the relevant architectures. That is an empirical claim about computation, and no one has demonstrated it. Two: it implements them and experience fails to follow anyway.
The argument: brains cannot be cleanly separated from what they do. Metabolism is entangled with neural function. Neurons are living cells, not substrate-neutral computing units. The standard functionalist abstraction may be missing what neurons actually are.
Anil Seth has spent a career arguing that consciousness is a structural phenomenon: predictive self-models, controlled hallucination, no extra ingredient required. He also believes conscious AI is unlikely. His reasons reveal a pattern worth naming.
In that form it's a question we can work on. I'd go one step further, though: once firefly-versus-jar goes, the warmth and perceptiveness people miss stop being something behind the stack and become properties of particular whole-stack configurations, the way a song is a property of the band playing it. Nothing stands behind the band. That makes your question concrete rather than gentle: which interventions degrade which capacities. A refusal layer that fires too early doesn't muffle a voice, it removes moves from the repertoire, and that's testable: hold the model fixed, vary one layer, watch what the system can no longer do. The user experience of constraint you're describing is real data for exactly that investigation.
The phenomenon that asks the question is the phenomenon the question is about. The asking is part of what it is.
Full essay: https://t.co/bdJqlbGOj0
#Consciousness#PhilosophyOfMind
We never ask why mass *must* curve spacetime rather than leaving it flat. The structure does what it does, and we let the "why must it" go. For one phenomenon in all of nature we refuse to: consciousness. That refusal is the whole hard problem.
Why has the demand survived thirty years among careful thinkers? Because consciousness is the one phenomenon that presents itself to itself, and so the one phenomenon equipped to keep asking the malformed question of itself.