Exploring State-based vs. Path-based Active Inference

This chapter delves into the distinction between state-based and path-based active inference, focusing on the stationary distribution of probabilities over states versus probabilities over trajectories. The discussion delves into different types of states within the Markov blanket, explores the physics behind the free energy principle, and addresses the challenges and implications of applying gradient descent and free energy minimization to internal and blanket states. The conversation also touches on the concept of self-representation, the role of collective behavior in Bayesian beliefs, and the complexities of incorporating predictive models in understanding psychopathological conditions like schizophrenia.