The Thousand Brains Theory

The dominant image of the brain in both neuroscience and artificial intelligence is hierarchical: information flows upward through stages, each processing more abstract features than the last, until recognition or understanding is achieved at the apex. Jeff Hawkins’s Thousand Brains theory overturns this image at its foundation. The neocortex’s approximately 150,000 cortical columns are not stages in a processing pipeline. Each one is a complete, independent sensory-motor modeling system—a little brain that builds models of entire objects, assigns features to stable locations in reference frames, and participates in the consensus that constitutes perception and understanding. The unit of intelligence in the cortex is not the neuron and not the whole brain but the column, repeated by the hundreds of thousands, each modeling the world from its own partial vantage and voting with the others on what is there. Intelligent behavior, on this account, is the emergent result of