Interesting, but after a quick read I don't see the paper addressing the possibility that the dynamical states can be first simulated by machines and that an AI consciousness would be embedded in such simulation. The paper only addresses AI architectures built directly on hardware and makes statements on the differences between computer hardware and the brain, but computer hardware could, in theory, simulate the substrate in which brains have evolved. Although this would probably be computationally too expensive with today's machines, so the paper's conclusion is still applicable today for that reason.