Seems that Seth doesn’t understand the Zombie argument at all. Assuming Seth believes in the causal closure of the physical world (I don’t think he believes consciousness is an immaterial force “filling in” the causal gaps of indeterminate physical processes in the brain), he should take Zombies more seriously. The Zombie argument applies to any physical process no matter how “complex” since physical processes can always be conceived to happen exactly the same way “in the dark”, as a zombie. If the physical world is causally closed, all the causal “work” is done physically in the brain in a coherent, intelligible way and consciousness is only assumed because we know about it from first person experience. Zombies are a convincing way to make the Hard Problem explicit via a thought experiment. His example of imagining a A380 moving backwards is irrelevant because the incoherency there is implied by the non-controversial ontological character of the matter that constitutes it: given that matter is what is and if when I’m imagining a A380 I’m really imagining a physical object, then I can’t “actually” imagine it moving backward because it wouldn’t really be actual matter I’m imagining. (What I could imagine is the phenomenal experience of seeing something “like” that happen, like a special effect in a movie. I’m actually imagining a potential possible experience.) Zombies are a different kind of conceivability question altogether. It’s precisely consciousness’ radically different ontological nature that the Zombie argument is attempting to bring to fore. To argue against zombies you’d have to demonstrate why physical processes MUST be conscious, (probably impossible given the fundamental modality of “physical” explanation itself) or introduce a new fundamental ontology of the world such that zombies are impossible because the concept of the physical world, as implied by zombie dualism, doesn’t exist. (IIT actually veers in this direction.)
Also, Phi is not at all poorly defined. You can analyze any system, find the spatio-temporal scale at which that system is most integrated, (the scale at which the behavior of the system is more than the sum of its parts and therefore fully analyzable only as a single whole), and then calculate using either the Kullback-Leibler divergence or the Earth Mover’s Distance (Wasserstein’s metric) - (different versions of the theory use different statistical methods) - the exact value of Phi as a measure of the amount of integrated information in a system. The fact that the theory is being “refined” is a testimony to its appeal. The original paper “A Provisional Manifesto” is still the best overall description of the theory, and one that actually goes into some (very complicated) detail about how the “architecture” of the n-dimensional information space implied by the theory maps onto phenomenology.
FWIW, I found Scott Aaronson’s analysis of of IIT intellectually uncharitable. He didn’t seem interested in understanding the theory on it’s own terms, and his main criticism, that a grid of logic gates could be conscious if IIT is correct, unconvincing. If neurons could be conscious, why not grids? There is even some neurological evidence that the brain is actually organized in grid structures, especially cortex, but the tangled and coiled up physiology of the brain obscures this fact.