Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

all right welcome everyone it's cohort 6 and our second discussion of chapter one so Andrew and Ali thanks for facilitating just take it from here wherever you want to pick up Andrew would you mind taking the lead on this one sure yeah so I know um yeah absolutely so welcome cohort 6 this is week two that we're spending on chapter one of the textbook um I'm not sure what was discussed at the last meeting but I thought it might be worthwhile for some of the folks who are unable to attend the morning meetings to maybe just quickly move through um the sections of textbook um just to quickly get a bit of a recap or overview um yeah the overview we begin with a nice brief introduction um describing what's going to be happening in this chapter we already get off the bat that active inference is a normative perspective it's starting from first principles um move on to section 1.2 get some quick explanations regarding the relationships between things like actions and sensory observations specifically that living organisms can only maintain their bodily Integrity by exerting adaptive control over the action perception Loop this means acting to solicit sensory observations that either correspond to desired outcomes or goals such as the sensations that accompany secure nutrients and shelter for simple organisms or friends in jobs for more complex ones or help in making sense of world um this is carried out through there's a rather recursive nature to this cycle of action uh perception Loops um also whenever we move on to looking at more complex organisms there's the assumption that organisms can increasingly use like more complex adaptive strategies to adjust to their environment um section 1.3 still more on um first principles get an introduction to this uh this kind of I don't want to call it ad hoc because I think these conversations have been had before but it's a a little light-hearted I suppose this meets versus scruffy's uh division that's main right and so um scruffies being like these are folks who will approach these kinds of problems but maybe from within their particular field or subfield um they don't want to um you know forego any kind of kind of idiocy idiosyncrasies or specificities to their object of study um so maybe you know not so quick to jump to some kind of broader unifying framework um to analyze things whenever you know you might let go of some of the specific the problem you're looking at um you anthropologists will have disagreements with sociologists and so on about what what defines like a society or or a culture or otherwise um nonetheless meats are those who do seek unification Beyond heterogeneity of brain and mind phenomena and that is the perspective that active in for instincts it does in fact attempt a kind of unified framework at looking at diverse uh scenarios and organisms ranging from human beings and entire cultures uh who are operating at different time scales of the way at fast time scales of like what they're immediately observing and perceiving um all the way up to slower time scales of like planning learning um organizing with a group development of Institutions and so on um section 1.4 gives us a good overview kind of structure of the book um 1.4.1 active inference in theory we get more about surprise minimization um we're also introduced to variational free energy in the next paragraph there um they keep it short in describing why we might look at variational free energy as opposed to surprise which seems like a much more intuitive concept um for one to be surprised at unexpected uh phenomena or observations or incoming um sensory signals um but the the to put it quickly variational free energy is basically an upper bound ons surprisal that is done because computationally um we find this issue of intractability in computation when whenever you break down the math you find like oh in order to actually precisely compute surprisal you would need something like an entire history of an…