Book review: Being You

Being You by Anil Seth is "A New Science of Consciousness", if you believe the subtitle. As I was dumb enough to. However, it isn't really about that. It is more a tour of some recent ideas about consciousness, and some of his own ideas. None of these ideas (spoiler, sorry) approach the real problem, i.e. "what is consciousness and how does it work?".

Before I go further I should confess my biases. I have some interest in the subject, I know perfectly well that there is no coherent theory of C, and so I knew perfectly well that the book wasn't about to explain it². And, I'm generally unimpressed by most philosophers³ and similar attitude (example; another). If you'd like to read another pile of reviews, try Goodreads.

My main criticism of the book is that he is talking about things I'm not greatly interested in, or already know. But you may not regard that as any great flaw of the book itself, only of my relation to it. Enough caveats! On with the show...

Quite a bit of the first part is about perception; in the sense of how does it work. This is interesting, in a way, but the connection to C remains a touch vague. He makes a decent case for the idea, which in retrospect I think I might consider obvious, that "ourselves" do not perceive the external world directly, instead we effectively sip from the top of a continually-updated model of the world. He has some less convincing ideas about error-minimisation as part of the feedback loops in the model, and he decides to call the models "hallucinations", but I think those are unimportant details so shall ignore them. What's less clear is if he regards the conscious-self as including the model, or whether the model is, in some sense, an extension of organs like the eyes, and merely constructs a model to be perceived by the true self. Perhaps he doesn't know. I think I don't know; nor do I know a way that you might tell the difference.

[Update 2023/03: I find The brain creates a predictive model. This just means that the brain continuously predicts what its inputs will be. Prediction isn’t something that the brain does every now and then; it is an intrinsic property that never stops, and it serves an essential role in learning. When the brain’s predictions are verified, that means the brain’s model of the world is accurate. A mis-prediction causes you to attend to the error and update the model from https://stratechery.com/2023/chatgpt-learns-computing/ as Jeff Hawkins's theory. This is similar, but different.]

At one point he discusses "internal" signals - body temperature, heartrate - as being folded into this model. I think he rather sneaks this in; it isn't clear they are part of the same model at all; and since they aren't under conscious control, it isn't clear they are part of C either.

There are some bits that seem to me like filler or padding. For example, he discusses the "but what if you could do Star-Trek style matter-transmitters, and copy people, are the copied people continuous with the originals?". There are problems with this: we can't do it, and I think QM says we cannot do it even in principle¹. And so the problem with discussing it isn't that we don't learn anything from it.

There's blobs of stuff about IIT and FEP but those aren't the answer either as I think he knows.

There's a chapter on animal C, but he has nothing to say on the subject; ditto one on machine C.

I think I'd recommend reading Godel, Escher, Bach instead.

Update: I should give him credit for physicalism: that C is a purely material behaviour; whatever it is, it arises from the substrate of atoms etc. that is our brain and body; there is no Cartesian dualism or soul involved. OTOH, for him it is only a minimal credit, since believing anything else would ruin his research programme.

Notes

1. Yes, I haven't thought this through carefully. QM only forbids-in-principle if you need to copy down to the finest granularity. Probably, you do; certainly, you do, unless you have a clear theory that says otherwise, and we don't. See-also me over there.

2. Or if it did, it would be wrong :-)

3. I know, he isn't a P. But you can't talk about C a lot without getting dangerously close.

Aside

Asimov, in e.g. Robots of Dawn has his more advanced Positronic robot brains in "humaniform" robot bodies: robots that look human, that have human-ish skin, the ability to smile, and so on. I am doubtful that this makes any sense, though it suits the storyline. He explains it as "And if I succeeded in working out a theoretical structure that would imply a humaniform positronic brain, I would need a humaniform body to place it in. The brain does not exist by itself, you understand. It interacts with the body, so that a humaniform brain in a nonhumaniform body would become, to an extent, itself nonhuman". It is gumpf, but perhaps interesting in this context. And anway gets contradicted by the ending.

Refs

* Blindsight as a phenomenon points to the... interactions of our wetware and conciousness.