Why the Brain Doesn’t Start From Scratch

Quick Overview

The brain learns new behaviors efficiently by composing previously learned, reusable neural modules, as demonstrated by monkeys quickly adapting to new tasks that required combining existing color and shape processing pathways, rather than learning from scratch.

Key Points: The brain learns new tasks by composing existing, reusable neural modules, a concept illustrated using a task where monkeys had to switch between reporting shape (Axis 1) and color (Axis 2) information. When the task switched from C1 (color-based movement) to S1 (shape-based movement), the color subspace in the prefrontal cortex was suppressed while the shape subspace remained active, showing context-dependent modulation. The researchers found that the neural activity representing the 'color' information (the color subspace) was physically reused in both the color task (C2) and the shape task (S1), indicating shared neural building blocks. The task belief decoder, trained on the color task (C2), could successfully predict the task context (C1 or C2) from neural activity, demonstrating that task context is explicitly encoded in the neural state. When the monkey performed a novel task (C1, color task) after training on C2 (color-based movement) and S1 (shape-based movement), the neural activity rapidly reconfigured the existing modules to perform the new task, showing compositionality. The speed of adaptation to new tasks is rapid because the brain is essentially plugging together pre-existing, task-relevant neural modules rather than creating new hardware from scratch.

Context: This video explains the concept of 'compositionality' in neuroscience, drawing from a 2024 Nature paper by Tafazolli et al. studying how macaque monkeys learn and switch between tasks involving visual discrimination of shape and color, and the subsequent motor responses. The core question addressed is why biological brains are so adept at learning new, related tasks quickly without starting the learning process entirely over.