Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

all right welcome back cohort 6 we're in our first discussion on chapter 5 so let's jump over there um where does anyone want to begin could be like any page or figure or quote of five that they remembered or that they're curious about so first we'll just see any thoughts or ideas people have about five otherwise we can look to the chapter and the prior questions sure I I thought that um this was a cool chapter because it for the first time I think kind of gets a lot of the message passing mapped on to um specific biological systems that we actually have quite a bit of prior knowledge about um so I thought that was pretty neat like to actually see the sort of detailed example of the Pathways in the basil ganglia and that kind of thing I thought that was great yeah it's go yeah I was taking a look at that myself I don't know if we could go through that a little bit because I was just having some trouble with the path tracing it might have just been the model itself and the way that I was misreading it but uh uh figure 51 and yeah 5 five is like the big overview abstract it's gonna it covers the prefrontal cortex graph the dopaminergic the basil ganglia graph and the spinal reflex arc graph and then earlier in the chapter so now we'll go back to 51 and so on earlier it's going to go through those examples and then those examples have a composition it internally that's the graph and then this is kind of the cool thing in the interoperability is that because of that compositionality internally you can kind of do wiring across graphs so that also speaks to the compositionality um of the models and I think maybe one theme that we'll try to draw out and and explore and see limitations of is like there's a massive underlying hypothesis of computational Neuroscience which is like computational models can map biological territories so there will be like some regions or some actual tissues that either do that's more of the realist angle or can be modeled as doing that's more of the instrumentalist angle they can be modeled as doing certain computations and if that's a viable hypothesis then the maps that we make with the math that we have are going to be very apt if it's an inviable hypothesis it could be extremely misleading because there could be a tissue and potentially the toolkit of mathematical operators just isn't adequate to make a good map of what that tissue does leading to like a false confidence about um the understanding of of of what it's doing in the body or like its role in development Evolution all of that so so it's kind of like it but that's so sublimated into the field decades on that it's like of course we're putting mathematical symbols on top of pictures of tissues but yet that's actually like that's kind of the invisible elephant in the room is like what what has happened here applying these mathematical models the graphical models and like using them as overlays to make sense of the functions of different tissues yeah that's that's really interesting I'd Wonder um so perhaps kind of like in addition to that one of the things that I had heard a few neuroscientists raise is the question of whether the um whether the kind of known connectivity maps in the brain suit themselves to this kind of explanation as well um oh I'm sorry Andrew you have your hand raised would you like to would you like to go oh yeah uh thanks um yeah no it was just a quick comment following from kind of what Daniel is describing this sort of like mapping the physiological to the to the comput or mathematical I suppose um I heard this nice explanation uh it's in a previous uh live stream with the Institute with uh Ryan Smith the computational psychiatrist and um it was just whenever I was watching it it was like my first time hearing like this kind of attempt to directly relate neural activity to to like I guess the mathematical equation so he was like breaking down computation of like variational free energy and expected free energy and like…