[The Cerebrum 6.3 : What it does] [Part 1: Why "Perceiving" is Actually "Prediction"] For decades, the standart model suggested that the brain worked as "bottom-up-inputs-video-camera"— the outside world pushing data into a passive brain. Now, thanks to Karl Friston and Andy Clark, we see the brain as the one who pushes a simulation out into the world. This suggests that we are seeing the brain's best guess about the world, a top-down simulation that is corrected by the bottom-up sensory data when it encounters novelty (Sensory prediction error) The Mechanism of the Predictive Loop Step 1: Higher-order areas send "predictions" down to lower-order areas The prefrontal cortex or association areas are essentially saying to the primary visual cortex V1: "Based on the fact that we are in a garden, these shapes on the grass are likely the hose" Step 2: Sensory Sampling (Bottom-Up) Photons hit the retina. This raw data travels through the thalamic LGN gate— Lateral Geniculate Nucleus and arrives at the cortex. [Precision-Weighting Signal and High Signal-to-Noise Ratio] However, cortex sends significantly more axons back down to the LGN than it receives from it. If the cortex predicts the hose (noise) with 99% certainly, it "tells" Thalamus to suppress their "sensory delivery" (potential signal) It ensures that energy is not waste lost on unnecessary sensory processing, opening the gate only for "a snake" (signal) sitting in the grass —a Prediction Error forces the update of vision. The "quieting" of the background makes the "news" more clear. Step 3: The Prediction Error The cortex compares the incoming sensory data with the top-down prediction. If the data matches the prediction (the shape is the hose), the cortex actually suppresses the signal. Cortex doesn't need to waste energy processing what it already predicted. However, if there is a mismatch (the "hose" is actually a snake), a "Prediction Error" signal forces the step four. Step 4: Model Update This error signal travels back up, forcing the higher-level models to revise their simulation. The "illusion" is updated to match the reality of the snake. This is a huge metabolic optimization. we only process the news—the parts of the world that didn't fit our internal model.
