Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

Okay, hello everyone. It's October 7th, 2022. We're in the sixth meeting for the second cohort of the textbook group. Today we're going to be discussing chapter three for the first time. So we have a few questions to explore on three. But before we do, would anyone like to just say hello or give any overview thoughts on chapter three before we go into more specific questions? You can raise your hand or just unmute and just go for it. I'll wait for someone to make a comment on chapter three. Thank you. Yep. Two active inference with the principle of least action and the introduction of Markov blankets. Okay. How does that differ from previous approaches or relative to what? Well, I guess instead of starting with the notion that you have some agent that's creating a representation of its world, you start with like an axiomatic view from physics that when applied to biological entities, leads to at least the same conclusion as chapter two. All right. Awesome. Thank you. Anyone can raise their hand or just go for it. General thoughts on chapter three and or what is the high road to active inference? Thank you. Having read chapters two and three. How does anybody want to remark on figure 1.2? Thank you. Okay. I'll make a short remark. To me, this figure, I guess, these elements like surprise minimization, Markov blanket, it looks like they are inserted in the pictures like sequentially, but it doesn't seem to me that they actually represent a sequence for understanding the approach. They're just various aspects of either the physics approach or in the lower part of the figure of the Bayesian approach. If I understood it correctly. If I understood it correctly. Because at first I thought it was like a path, you know, starting from the free energy principle, then first to self-organization, then surprise minimization, Markov blanket, predictive processing. And I don't know if it's, if it is actually a progression of concepts or actually, I don't know. What do you think? Yeah. It's a great question. Anyone can give a thought. What do the ordering of the terms mean? Are these train stops along the way blue or child? So I just wanted to say that the live stream where Maria really gave a lot of background on predictive processing, predictive coding, like the difference between these two might be super helpful with regard to like a progression or just evaluating that lineage in more detail. Yeah, sure. I'll add link there. One option would be that these are historical. Like this is this chronology of ideas. Another one would be that they're pedagogical, that this is a learning path. Another might be formal or technical, like one of them are one of the directions are dependencies or required, but not necessarily the most pedagogical path. Ali? I also think that the organization of this particular figure is more like the branch of a tree than something sequential. So they might not represent any specific preference to go through all of these steps in order to achieve a particular goal. They're just try it just tries to lump together relevant concepts in two different branches. So that's what I see it at least. That sounds weird. That sounds weird. That sounds weird. That sounds weird. That sounds weird. That sounds weird. That sounds weird. That sounds weird. That sounds weird. from 2010 with a lot of the early sort of directions and pointers towards areas of synthesis. Anyone else want to just add any other thoughts like what is the high road or how did they feel about chapter three overall? There's ample technical and specific questions to dive into, but it's really important to understand chapter two and the low road and chapter three being the high road and the way that they're distinguished in terms of the low road beginning from a framework for doing statistics, learning, updating, and so on, but not necessarily an imperative for survival. In contrast, in the high road, the beginning stance is around the imperative for survival,…