The Thousand Brains Theory of Intelligence, proposed by Jeff Hawkins and his team at Numenta, suggests that the neocortex does not rely on a single, centralized model of the world but instead builds thousands of individual, parallel models of objects and concepts simultaneously.6 Each cortical column, a fundamental unit of the neocortex, creates its own complete model based on sensory input and movement, using a reference frame to track objects in space and time.4 These models are not static; they are continuously updated through sensorimotor interaction, allowing for dynamic learning and adaptation.3

Rather than processing information through a strict hierarchical pathway, the theory posits that intelligence emerges from the collaboration of these thousands of independent cortical columns, which communicate and reach a consensus through a voting mechanism.4 This decentralized approach allows for greater resilience, as the loss of one region can be compensated by others, and enables robustness against failure.4 The theory is grounded in neuroscientific evidence, particularly the role of grid cells—neurons that represent location in space—which are found not only in navigation centers but also throughout the neocortex, supporting the idea that the same mechanism is used to learn the structure of objects and environments.2

The theory challenges traditional artificial intelligence, which relies on deep learning architectures that process information through a single, hierarchical pathway and require vast amounts of data.3 In contrast, the Thousand Brains Theory suggests that AI systems could achieve more human-like intelligence by mimicking the brain’s distributed, parallel modeling approach, enabling continuous learning, reduced data requirements, and better generalization across domains.3 This framework has inspired practical implementations such as "Monty," a system developed under the Thousand Brains Project, which aims to build AI applications with enhanced sensorimotor intelligence and adaptability.9

The theory also provides insights into higher cognitive functions, including self-consciousness and memory, proposing that personal identity and recollection emerge from the interactions between cortical columns and the distributed storage of knowledge within reference frames.4 Hawkins argues that current AI systems lack true intelligence because they do not possess reference frames or the ability to model the world dynamically, and he envisions future machines capable of human-like learning and reasoning.10 The theory has been detailed in Hawkins’ book A Thousand Brains, which outlines its implications for neuroscience, artificial intelligence, and the future of human-machine collaboration.6