Anil Seth just described the free energy principle as "largely impenetrable" 🤣.
@DrYohanJohn Ha! It is pretty opaque. Friston and co have tried to boil it down into an intuitive and easily digestible form, but the details are pretty convoluted. This makes me want to give it a try.
I've spent some time on it, going through the math in great detail, and I've come to the conclusion that it's a waste of my time. The notation is a shambles, the physics connections questionable, and the conclusions vague/untestable. What is true in it strikes me as trivially so.
I'll wait for a textbook entry five years from now. 😅
@DrYohanJohn Understandable! Friston tends to the grandiose, which has made me wonder how much is puffery (hanging mostly on a high-level similarity between non-equilibrium thermodynamics and learning) and how much is genuinely helpful. There are a few bits I have gleaned from active inference that seem useful in my context, such as the significance of precision-adjusted errors in feedback, but none of it hinges on his grand approach.
@johnnylogic
Friston juggles three concepts. Here's how I rank them:
1. Predictive coding (some form of this seems obviously true)
2. Bayesian networks (not a fan myself, but this framework is not obviously wrong)
3. Free Energy Principle (barely coherent in my opinion)
(Also, he only came up with #3.)