Ok so I'm gonna try to tie #2 and #3 together. This is not how I'd try to convince someone who starts out skeptical but it's more constructive than just focusing on paradoxes.
I think the best starting point is thinking about computation, specifically what you are using to compute. If physicalism is true then all computation must be based on physical mechanisms in the broadest sense, and if so, we can roughly differentiate between physical complexity, which is how complex of a mechanism you use, and logical complexity, which is how many of your mechanisms you are chaining together. I've made this graphic for an article a few weeks ago that illustrates this:
So one extreme case is what I call standard computation, which is where you use ultra primitive mechanisms but you go apeshit in how complex your logical arrangement of those is. All digital computers (and actually analog computers as well) fall into that category. Your only mechanisms are two different voltage levels to differentiate two states, and some wire designs to implement binary logic on those two states. That is almost literally it (I think some marginally more complex physical pieces exist in some computers). Standard computation is also the only thing that's analyzed under "theory of computation" in mathematics. Like computability, turing machines, complexity theory, all of that stuff assumes that you have a very basic set of operations that you can combine in arbitrarily complex ways.
Then nonstadard computation is just everything else. Clever mechanical divices for integration, a soap bubble that "computes" a spherical shape, a wire whose surface you make repulsive such that it self-untangles, clever arrangements of coupled oscillators, and also the EM field.
My claim then would be that
standard computation cannot support consciousness. It doesn't matter whether it's analog or digital or based on electricity or water; the important feature is that you have a set of extremely primitive physical mechanisms and all the complexity comes in at how you chain them together. And the way to see that this doesn't support consciousness is just that it fundamentally doesn't have the right properties.
So for example, consciousness seems to have objective, discrete boundaries. There's a set of qualia in your consciousness, and a different set in mine, and they're perfectly split. If we start interacting, they remain perfectly split. That gives you two paradoxes, one is that no qualia sharing takes place even though we causally interact, and the other more fundamental one is that it doesn't even seem logically coherent for partial merging. Like, any quale can either belong to one set or the other and not both, but the degree of functional integration between systems seems to be a spectrum. (Btw, qualia are just literally all components of your consciousness, tactice, visual, etc; anything and everything you perceive is a quale or set of qualia.) This is called the boundary problem, or binding problem, or combination problem.
Another issue is that consciousness seems to have inherent content. Like, red looks a certain way, a harmonic chord sounds a certain way, and more subtly, all qualia seem to be valenced (i.e., feel good or bad). You can't bootstrap yourself toward that with standard computation because you start from empty content (just two abstract states) and you can't ever get from there to nonempty content. (I think this argument is very unpersuasive to people not already sympathetic to this idea, but I do think it's true.)
Yet another problem is that standard computation, since it's entirely based on abstract states, is fundamentally dependent on interpretation. Like, suppose a digital computer were conscious and right now would experience pain. All that's actually happening at the physical level is that data strings are moved around (i.e., sequences of bits, which physically are sets of pieces on a wire that either have around 1V or 2V). But presumably the consciousness (so the felt pain in this case) is frame-invariant (i.e., it is one way regardless of who interprets it). So how does that work? Does the universe try to interpret what's going on by looking at the surrounding system? That doesn't sound like something that could be an actual law. Does it depend on the system's counterfactual response to other inputs? If so then you have more to work with, but now consciousness depends on counterfactuals, which physics does not, so you lose connection between consciousness and causality. Mike Johnson had this thought experiment where he just said, suppose I shake a bag of popcorn, did I just torture someone? It's difficult to argue that the answer is no since there would be an interpretation of the atoms in the bag and how they correspond to informational states that makes the computation represent the simulation of someone who is experiencing pain. So this whole strand of argument is called the interpretation problem and seems altogether impossible to resolve.
I think the
underlying problem behind all this is that small discrete objects are just the wrong level of interpretation. If consciousness is supposed to be operating within the laws of physics, then there must be a deep connection between the laws of physics and the laws of consciousness. In fact they should be isomorphic, i.e., equivalent in a deep mathematical sense. Certainly what they say about causality should be exactly equivalent. But the laws of physics are written over fundamental forces and fields, so the laws of consciousness must be as well. That means any approach that first tries to round everything to an ontology of discrete computational elements like logical 1s or 0s (or rational numbers in case of analog computers) is just inherently doomed, and all the problems above and various others are just ways at pointing at consequences of this disconnect. You have to analyze consciousness at the lowest possible level, the same as physics.
And well then the EM field is pretty much the only candidate. It's one of the fundamental fields in the standard model, and all of the others seem like non-starters. It solves the interpretation problem because it deals in physical forces rather than abstract codes that need to be interpreted, and there is this physical phenomenon where the field lines can form a topologically closed shape that makes it causally separate from the rest of the field, which solves the boundary problem.
)