Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

Hello, this is Active Inference Textbook Group, Cohort 2. We're in Meeting 10 on November 4th, 22, and we're having our first discussion on Chapter 5, Message Passing and Neurobiology. So, to begin, does anyone want to share any general thoughts or remarks on Chapter 5? Just any experiences they had reading it? Any topics that they thought they want to discuss a little bit today? Just anything about their reading of Chapter 5? Yes, Ali? Well, I think in Chapter 5, since the authors decided to go pretty deep into neuroanatomical structures or neuroanatomical substrates of most of the Active Inference concepts, well, having a kind of basic understanding of neuroanatomy, neuroanatomy, in my view, is an essential prerequisite to understanding the contents of this chapter. But otherwise, it provides some empirical evidence for why Active Inference Modeling works, or to put it this way, it explains some of the testable predictions we can get from applying, from seeing Active Inference from a neuroanatomical perspective. Awesome. Thank you. Anyone else want to share a general thought on Chapter 5? Yeah. As Ali stated, the chapter is going to provide a host of examples, which are going to be discussed separately, and then synthesized together, regarding neurobiology, specifically mammalian neurobiology, which is one of the most important systems of interest of us as humans. And also, it's one of the most studied in terms of different zones where Active Inference models have been applied, ranging from, as we'll see in this chapter, the micro circuit level. So, the connectivity patterns of small numbers of computational elements, which might be like neurons, or it could be brain regions. Then, there's a discussion of motor commands in terms of continuous time Active Inference, and how motor behaviors in the periphery are coordinated. There are then several discussions about decision-making, neuromodulation, and learning. And then, there's a hierarchical hybrid model that's presented, which integrates the continuous and the discrete cases of Active Inference in a unified generative model. And that's what brings it all together in the summary. And also, just one note, maybe we've mentioned before, Chapter 1 provides an overview of the book, much like Chapter 10. Chapter 2 and 3 are the roads to Active Inference, but not Active Inference itself, though it also kind of is. Chapter 4, we learned about the essence of specifying an Active Inference entity, which is its definition in terms of a generative model. And now, immediately, we're moving to the area with the most empirical support and popular interest, which is understanding Active Inference in the context of mammalian neuroanatomy. And that will conclude the first half of the book, which is the epistemic component. The second half of the book, Chapter 6 through 10, heads into some of the step-by-step and recipes around designing Active Inference models. But in some ways, as we complete this last chapter in the first half of our cohort, it's just good to keep in mind that we spent 60% of the time getting into the game. Then, where the low and the high road intersected was in the generative model. And now we're immediately going to be talking about topics related to applications and specific generative models applied to neurobiological phenomena in health and in disease. Okay. Does anyone else want to add a comment on Chapter 5 before we start moving through it? Okay. Does anyone else want to add a comment on Chapter 5 before we start moving through it? All right. Chapter 4 described the generative model. In this chapter, there's a focus on the process theories arising from those dynamics, theories that explain how the brain may implement variational inference. So, as per classical scientific criterion, the goal of a theory or a contribution is to variously explain, predict, design, control. These are the criteria that active inference can and should be compared and juxtaposed with…