6 ideas
The brain does not passively read out sensory data; it continuously generates top-down predictions about the causes of its sensory signals, and what we experience is the content of those predictions. Incoming sensory signals act mainly as prediction errors that correct and constrain the guesses, so perception is a hallucination held in check by the world rather than a window onto it.
The deepest layer of the self is the brain's prediction of the body's internal state (heartbeat, blood chemistry, gut), carried out for the purpose of keeping physiological variables within survivable bounds. Emotions and moods are the felt experience of these interoceptive predictions, so being a self is fundamentally about regulating a living body, not about reasoning or representing the outside world.
Instead of asking why physical processes produce experience at all, the 'real problem' approach asks how to explain, predict, and control specific properties of consciousness in terms of brain mechanisms. Consciousness is split into level (how conscious), content (what one is conscious of), and self (the experience of being someone), and each is mapped to mechanisms, on the bet that the metaphysical mystery shrinks as explanatory bridges accumulate, as happened with 'life' in biology.
The self is not a fixed entity doing the perceiving but is itself a set of perceptions constructed by the same predictive machinery: bodily ownership, first-person perspective, volition, narrative identity, and social self. Because each layer is a prediction, each can come apart separately, as in rubber-hand illusions, out-of-body experiences, or depersonalization, revealing the self as assembled rather than given.
Seth inverts Descartes' picture of animals as unfeeling mechanisms and proposes that our conscious experience may be bound up with our being living, self-maintaining organisms. He offers this as a direction for explanation, suggesting consciousness may have more to do with being alive than with being intelligent, not as an account of how self-maintenance gives rise to experience.
Seth argues that intelligence and consciousness are distinct, so AI systems could become more capable without thereby becoming conscious. He suspects that being a living system may matter for consciousness but presents this as an intuition rather than an established result. He also flags the risk of machines that merely seem conscious, which could exploit our tendency to attribute minds and distort our moral judgments.