Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

all right greetings everyone it is October 10th 1010 23 and we're in our discussion of chapter 7 so last week Ali and I recorded like the short video overviews of chapters nine and 10 just because there was um only a few of us during our cohort for slot and we just wanted to complete those um overviews so now we're in the chapter 7 um discussion so we'll jump over to the chapter 7 page um and before we look at like any of the specific questions like does anybody just want to share um what did they think about chapter 7 what what was there impression or or how would they summarize it to to somebody or what was just exciting to them about chapter 7 well can I start um so please hello Daniel hi yeah go forward go forward yeah so so my um impression of chapter 7 is about uh uh it's many about about the uh discrete uh State and discrete outcome model of the active inference and I found it is particularly uh useful to have a detail understanding of one of the example which is this T's uh example and it is kind of the first uh active inference model that I can get my hand on and try it out and change those uh those parameters or the the matrixes used in the active inference model and see how um how the uh active inference uh actually works awesome I think that's kind of the point um anyone else want to give a thought on seven um yeah I would just second everything that uh Lan said just now um I think it's a great way to actually see maybe how some of the the ideas that we've been reading about prior in the textbook actually start to hang together and play out in an example I think that the the use of discret variables is highly like kind of interpretable I think this is a it comes a little bit more naturally to me to read this chapter versus um you know chapter 8 uh with continuous variables because we get to see like your you know more precise specific things that happen discreet choices and and things that occur and then also just some kind of interesting observations that come out of that with the different um kind of components of the model like uh you know as you get further in the chapter there's a discussion of like here are the interesting things that happen whenever you don't include preferences or have highly uh uniform preferences like there's just pure novelty seeking on the part of the agent who like um like in the team's example they just can't sit still they never come back to the same spot they just it's pure just seeking epistemic value it seems because there's nothing pragmatic to do because there's nothing in particular that they're going for because it's not been hardcoded into the initialized um preferences so so yeah just um giving us that that really big model for active inference here in the discret form in addition seeing the interesting dynamics that can kind of play out emerging properties and so on awesome anyone else want to make a general comment on seven I would I would just Echo that given that I've come from a psychology background and so seeing the exploration exploitation Paradigm just being laid out in that way and they make the comment that you know they believe they've uh come to understand how these two go hand in hand instead of being competitive I thought that was also very very intuitive cool yes yeah the continuous time setting is kind of like having your hand on the rudder in a boat and uh so you're making continuous actions and there's a continuous flow in these paths and the discreete time is like chess or Checkers like time clicks discreetly and the actions are discreet the states are discreet and so that can make it kind of clearer a little bit more cut and dry um cool does anybody want to jump to a specific section of seven or we can look at the questions and just like look at at the upvoted ones or if anybody submitted any and just like look at what's interesting in seven or or somebody can propose a section that that they want to look at I'm yeah I'm mostly interest see in chapter…