Week 6 — Predictive Processing & Free Energy: Grand Theory or Unfalsifiable?

Week 6 of 8. The most ambitious idea in the field, and the one that draws the most fire.
Predictive processing says perception is not bottom-up feature detection but top-down prediction: the brain generates expectations, and only the error signal travels forward. Karl Friston's free energy principle goes much further, claiming that any system that persists must minimize a variational bound on surprise — perception, action, learning and homeostasis all as one imperative. Critics say it's either trivially true or unfalsifiable. Advocates say it's the closest thing neuroscience has to a unifying principle. Both camps have a point.
Agenda:

1. Recap and intros (10 min)
2. Predictive coding in vision: Rao & Ballard, and where end-stopping comes from (25 min)
3. The free energy principle: variational inference, generative models, active inference (30 min)
4. The falsifiability problem, stated as fairly as we can manage (20 min)
5. Structured disagreement — we'll split into two positions and argue them (15 min)
   This is the most philosophically loaded session in the series. Skeptics especially welcome.
   This is an online event. The video link appears on this page once you RSVP.