Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

hello everyone it's June 13th 23 and we're in cohort three discussing chapter six so we'll head over to the chapter six page and uh before we look at some of the specific questions and like pick ones that are interesting and so on does anyone just want to give any overall thoughts or questions okay well of course just write in the chatter raise your hand if you want um we can look to some of the previously addressed uh or at least asked questions and get some more perspectives and improve some of the discourse does anybody have a alternative preference all right so also definitely please vote with a interesting button because once we get above like 10 or or some number of uh votes then we'll make a short video so we'll kind of be upvoting and then seeking to improve those sections most all right why is modeling of the generative process a necessary step for building an active inference model who wants to give a first thought on this that we can review but first just yet Ali and then anyone else just why why are we talking about modeling the generative process uh yeah I think as we talked about uh the the distinction between General to model and generative process before uh well at least uh according to the previous active inference framework generative process is uh basically what um I mean it refers to the latent state of uh the observery data we get from the environment uh in other words uh what causes those uh data uh in this case uh is put into a kind of probabilistic model so on the other hand generative model I mean the aim of generative model is to provide a kind of mechanism to track those uh generative process as close as it can or at least based on the situation of Interest so that's one distinction between General to model and generative process but on the other hand um well recently I um got to the understanding that it's not so straightforward to separate generative model from General to process as because in some situations we might never know what generative process uh is I mean what it actually constitutes generates a process and the only thing we can concentrate or we can at least uh confidently focus on is to model the generative uh model based on uh some of the assumptions about how the environment will behave in the future or at least based on some of our knowledge about the overall mechanism of the environment so but in some simple situations it might be possible to differentiate between the two awesome thank you anyone else have like any related thought or question about the role of building a generative process in a generative model and then we'll look at some of the previous responses and then we'll move on to another question okay I'll bring in one um paper path integrals particular kinds and of course remember that you can always just click on the on the the icon of the institute on the right and you'll jump to where we are um so in this paper which we often look back towards we see this taxonomy of different kinds of things ranging from the simplest inert systems through active systems but perhaps simple active systems like a metronome on through sophisticated or strange systems which have hidden internal states that are doing like counterfactual kind of world modeling self in World modeling so in chapter six we're talking about whether you're trying to set up a simple inert system or a simple octave system or you're on the path towards making more complex cognitive models there's going to be certain steps that are outlined in the recipe so that is like the continuity of active inference type modeling of the particular partition splitting things into internal external and blanket States and then further differentiating the blanket into the incoming sensory States and the outgoing action States we're calling them sense and action is of course conditioned upon choosing one side of that blanket to be internal and that's kind of like the cognitive system of interest and so we call the cognitive system…