Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

all right greetings everyone welcome back it's Halloween 23 and we're in our second discussion on chapter 3 so we will look at the cohort 5 questions on chapter 3 um so I encourage people to add even now questions directly to cohort 5 questions um or the or or or there's all this other questions so maybe people added directly to the table and so it doesn't have the cohort five tag but before we go to any questions does anyone just want to bring up some piece of chapter 3 that stuck with them or highlight some other topic or share a little bit about how they're seeing the low road and the high road now that we're getting to our second discussion on the high road yes Susan then anyone else who raises their hand so does the low road and the high road are they differentiated by active inference versus Active Learning no Active Learning was the term used by that one paper just to sort of spin the fast ball a little bit different to highlight kind of like a regime that an active inference agent could be in instead of a purely confir confirmatory sensory mode rather a mode that that reduces uncertainty maximally about world States but active inference is the bigger Circle which what the authors characterized as active learning is a smaller subset Within which incorporates it um and um so Dean um brought up that active inference was a process of provisioning can you speak to that we would have this was in private conversation so we'd have to know a little more about what he meant um but but it's a very evocative yeah so that but that that kind of brought the question of differentiating active learning from active inference versus belief updating so because learning would involve belief updating right yeah as as with a huge number of the topics that we hear about belief updating attention learning anticipation there's often a narrower technical sense like belief updating is just any basian parameter updating at all so belief updating doesn't always refer to a human with a psyche undergoing some kind of like narrative development it can be a lot more narrow and small and less loaded with like philosophical or cultural baggage than if we were to say Hey when's the last time you had a belief updating most people aren't going to say Well when one of my proprioceptors in my arm you know did something different um so sometimes yeah the narrow sense or the technical sense of a term and then the the broader um um scope of what it might evoke and and so these are all good comments um Kristoff then anyone else who raises their hand uh I I just want to say that the high road and and sort of the kind of um uh you know the road to active inance through this ideas of self-organization and self- evidencing that that really reminded me of um some work that I'm not familiar with uh deeply mathematically but conceptually seems very similar that's um German England's work on the thermodynamic origin of life he kind of um in his work he has shown that if you have um uh system that's receiving an energy from from external source and it has a place to dissipate that energy and there are certain other conditions that need to be met then the system will necessarily organize itself in such a way uh to take advantage of that energy source um so it will start creating structures that will allow it to um um to uh you know utilize that energy and in his talk I think that was at MIT a couple years ago you know at the very end of the talk very kind of he makes this offhand remark that probably um when the energy coming in has some kind of temporal structure that may eventually lead to the emergence of intelligence or something like that that kind of made me when when when he said that in talk made me think about you know faradise speculating about the electromagnetic nature of light uh he's he's probably right and I'm pretty sure that he's I'm quite sure because he started with thermodynamics with his work that it can be recog consed um with active inference so…