No Way Out Karl Friston Decodes the Real OODA Loop: Active Inference and What Boyd Got Right
36 snips
Apr 28, 2026 
Karl Friston, theoretical neuroscientist and architect of the Free Energy Principle, explains how living systems perceive, predict, and act under uncertainty. He unpacks Markov blankets, generative models as orientation, expected free energy as risk and ambiguity, links to flow and psychedelics, and why current large language models cannot truly orient.
AI Snips
Chapters
Books
Transcript
Episode notes
Generative Models Are Internal Hypothesis Engines
- A generative model is an internal probabilistic hypothesis about hidden causes that predicts sensory inputs and enables Bayesian belief updating.
- Friston shows even simple systems can be read as inferring causes, while higher organisms have hierarchical, self-modeling generative models that support planning.
Actively Sample Smart Data Not Big Data
- Use action to select the most informative sensory data rather than passively collecting big datasets.
- Friston recommends active vision and targeted sampling: move your eyes or instruments to resolve uncertainty and get 'smart data' not just more data.
Expected Free Energy Is Planning Math
- Expected free energy formalizes planning as minimizing risk and ambiguity and maximizing information gain when simulating futures.
- Friston maps risk to KL divergence between prior preferences and predicted outcomes, linking economics' risk concepts to active inference math.
