Parr, Pezzulo, Friston 2022 Textbook Cohort 5, Chapter 6 part 2

Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

Feb 12, 2024

▶ Watch on YouTube ↗

Session details

Date: Feb 12, 2024

Series: Parr, Pezzulo, Friston 2022 Textbook Cohort 5, Chapter 6 part 2

Paper: Active Inference: The Free Energy Principle in Mind, Brain, and Behavior

Transcript

AI-generated transcript excerpt

The full transcript is available on GitHub. This excerpt is generated by automated speech recognition and may contain errors.

all right welcome back everyone it's February 12th and second discussion of chapter 6 for cohor 5 so any thoughts or feelings on where to begin with chapter like any thought or quote or [Music] question I have one yes please um I'm not so sure it's in this chapter but can you go into more detail in terms of uh the difference between the high road and the low road yes would anyone like to give a thought on high road and low road in the chapter six so really getting to the recipe and the cooking what are the high road and the low road well one thought is chapter 6 is a recipe for actually walking on the low road because this is where the generative model is constructed and specified so it kind of starts from having nothing I don't know how deep we um uh have or or need to go into the food maybe it's somewhere here but I mean what are the ingredients what's the kitchen what needs to be done in what order that's the low road of building any model there could be a chapter six for a recipe for Designing SPM analysis or designing linear regression analysis as thousands of other guide books for modeling are out there so in that sense that all modeling and uh construction like this is low road because it's actually putting uh model down that's what's similar about chapter 6 and another recipe for basian modeling and then the difference with active inference is the connection through the high road to the free energy principle so that's in so relative to if I were to to translate that um and you know and I'm the model that I'm I'm building is more of a a self-modeling not trying to build a software llm model um so in terms of yeah the low the low road is is Imagining the pathway and the high road would be taking the pathway would that be a way to contrast them you know simp IC form what would someone else say Yes um Ali and then anyone else who raises their hand okay so uh the high road uh can we described kind of topown approach to active inference which starts by asking some uh very fundamental questions about uh about what's required for any Cod UN code thing to exist and what are its necessary properties so uh by beginning from uh a kind of as friston himself would call it a near existential nihilism uh it gradually begins to build uh some framework which can account for any sentient behavior that agent must possess in order uh in order for it to possess its integrity and to persist through time so that's basically uh the premise of high road but uh on the other hand for low road it begins by asking it it begins by the premise of uh the existence of the mind so here uh the the existence of the agent or rather the predictive agent is presupposed and then it begins to formulate or describe uh What uh kind of mechanisms uh can realistically describe uh its uh predictive behavior and it's uh I don't know brain activity uh or rather the way that the brain act the brain constructs uh the inner model or uh in know words its generative models uh so that's basically beginning from the premise of the existence of mind and uh to describe uh what kind of um mechanisms uh that particular mind should possess in order for it to uh to do both inference and action in a changing environment but ultimately they uh they achieve similar goal but uh as uh the I mean the name of these approaches would suggest uh only the approach to achieve that same goal uh is different but then again uh they basically try to uh I mean describe the same kind of behavior so you the near existential nihilism would uh would that um would that correlate with let's just assume they have agency okay so uh all right so yes agency is a kind of I mean Central presupposition for any kind of um any kind of thing that can meaningfully uh navigate through its environment and to have any kind of goal directed Behavior so in other words agency is a kind of synonym for a gold directed Behavior it can be um obviously I mean argued whether agency really equates uh gold directed Behavior…