Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

Hello, it's February 22nd, 2023, 222, 23, and we're in the second discussion of Chapter 7 in Cohort 2. Before we go to the questions of Chapter 7, and then at the end, a look ahead to Chapter 8, we are going to have Thomas Parr join for a live stream in five days. And so anybody, if they would like to raise an area of questions, could be conceptual, or it could just be a general textbook level or personal question. So of course, feel free to add anything, but just while we're here, any thought or like kind of general piece that people are thinking that Thomas could reduce our uncertainty about? All right. Just, of course, add it when you see fit. Okay. There were several questions that we've already looked at from 7 and several more. We can also look a little bit more. Just again, we looked at the first sections of the chapter itself. So, does anyone have a Chapter 7 question that they want to go to or add? Or do they prefer to walk through the sections in the textbook? Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Let's look at equation 7.4. See if we have anything annotated for it. Okay. Okay. We don't have the natural language yet. None of the natural language yet. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. Okay. So, let's go back up with the rat in the maze example. Again, go through it. Connect it to the equations and then return to several of the questions down here in 7. So in the earliest part of chapter 7 was a passive inference with the example of listening to music. Now the upstairs action selection part is going to come into play. The imperative for action selection is going to be G. G, like we saw in chapter 5, takes the action prior, the prior on different policies, which can be understood as like habit, and then sharpens it according to, just qualitatively, provide epistemic and pragmatic value, and quantitatively as described in equation 2.6 and 7.4. Does anyone want to give a reading or a thought on 7.4? Just what is it saying in and of itself? And or how does or might this relate to, in their words, the resolution of the exploration-exploitation dilemma? Yeah, Ali, go for it. Well, yeah, as we saw before in chapter 2, I guess, but I think it was in more comprehensive terms. But here we only have a part of that equation, which basically deals with the trade-off between negative, I mean, epistemic value and the pragmatic value, and how they relate to each other. Because as we saw in chapter 2, the trade-off or the dilemma between exploration and exploitation is basically one of the fundamental requirements for any sentient behavior, because as long as the agent tries to, wants to explore the environment in order to informally discover new, let's say, opportunities or new resources, be it, I don't know, food resource or any other resource necessary for its survival in the environment, it needs to somehow have a trade-off between this kind of having the pragmatic value on one hand, namely to survive in the environment, and on the other hand, to discover new resources for its survival, which basically relates to the pragmatic value. But either of those two requirements, if taken to its extreme far, would probably result in the demise of the agent, because, so take, for instance, an agent residing in an environment, but for a specific, a certain period of time, it can survive in that environment using the pre-known resources there. But when the resources is about to, is about to diminish, then that agent needs to explore a little bit further in order to discover some new resources and continue its persistent through time or its survival. Yeah. Great. Yes, Michael. So, I'm a little bit wondering, in this epistemic considerations, only things can be considered that the agent actually already knows about, right? And then I'm wondering why, at several places, I think I said this in chapter 2 and here in this chapter 2, also 7 too, also that one would like to find the most…