[Verse 1] Your cortex builds a model, layer by layer high Each level makes predictions for the one below to try When reality don't match what your neurons thought they'd see Prediction errors bubble up the hierarchy From V1 to the frontal lobe, a cascade of surprise The brain's best guess at what exists behind your eyes Not passive like a camera lens, but actively creating Every scene you think you see, your mind is generating [Chorus] Predictive processing, hierarchical dreams The Bayesian brain computing what everything means Top-down predictions, bottom-up corrections flow Only errors propagate, that's how the patterns grow Free energy minimized, uncertainty reduced Your world model updated, new knowledge introduced [Verse 2] Rao and Ballard showed us how the visual cortex works Not copying raw pixels, but finding where the quirks Between expected and observed create the learning signal Prior beliefs get updated when the math says to jiggle Knill and Pouget revealed the probabilistic dance Every perception weighted by statistical chance Your neural networks calculating likelihood and prior Building generative stories that never seem to tire [Chorus] Predictive processing, hierarchical dreams The Bayesian brain computing what everything means Top-down predictions, bottom-up corrections flow Only errors propagate, that's how the patterns grow Free energy minimized, uncertainty reduced Your world model updated, new knowledge introduced [Bridge] Friston took it further with his grand unifying scheme Free energy principle, the ultimate brain meme Perception, learning, action - all minimize surprise But critics say it's too broad, unfalsifiable disguise Where are the error units? Where predictions truly live? Compositionality missing from the framework that they give [Verse 3] An idea's a latent variable hidden in the code Reasoning adjusts the weights along the neural road Active inference drives you to sample and explore Reduce uncertainty about what the world has in store From sensory cortex up to abstract thought domains The same predictive logic coursing through your neural veins Hold this lens with careful hands, it's powerful but loose A research program, not a proof, for minds to reproduce [Verse 4] In autism and psychosis, prediction might run wild Or attention weights too fixed, like patterns fossilized Depression as high prior weight on negative belief While meditation trains the meta-cognitive motif Clark and Hohwy debate the bounds of this grand view Does prediction reach all minds or just a chosen few? [Chorus] Predictive processing, hierarchical dreams The Bayesian brain computing what everything means Top-down predictions, bottom-up corrections flow Only errors propagate, that's how the patterns grow Free energy minimized, uncertainty reduced Your world model updated, new knowledge introduced
← Module 6 — Cognitive maps: reasoning as navigation | Module 8 — Global access, workspace, and cognitive control →