A Thousand Brains

Hawkins suggests that the neocortex is not a group of specialized regions but a sheet of 'cortical columns,' each acting as a complete, small learning system. Each column, no matter its location (visual, auditory, or somatosensory cortex), uses the same basic method to learn, predict, and model the world. This sameness is important: it means the brain does not need to create new methods for new tasks; it just uses the same strong, general learning mechanism for all sensory inputs and motor outputs. This idea challenges the traditional...

A key idea is that the brain does not just store sensory images. Instead, it builds and uses 'reference frames' for every object and idea it encounters. Think of a cup: the brain does not just recognize 'cup'; it understands the spatial relationships of its handle, rim, and base relative to the cup itself. When you move your hand to grab it, your brain is not calculating exact coordinates; it is mapping your fingers' position relative to the cup's reference frame. This system allows for strong object recognition from any angle and hel...

The 'Thousand Brains' theory is the main point: each cortical column, with its reference frames, builds its own complete model of a specific object or idea. For example, when you see a coffee cup, thousands of columns might be modeling different parts of it at the same time — its color, handle, weight, and use. These independent models then 'vote' or share their findings to agree on a conclusion. This distributed, parallel processing makes the brain very robust (damage to one column does not break the whole system) and explains how we...

Hawkins argues that the brain's main job is not just to process static sensory input, but to predict what will happen next when we move. This applies to physical movement and also to 'mental' movement — shifting attention, remembering things, or thinking about ideas. When you 'think' about a cup, your brain simulates how your eyes would move to scan it, how your hand would feel it, and predicts the sequence of sensory inputs. This constant, active prediction, driven by an internal simulation of interaction, is the core of thinking. It...

Our lasting sense of self, the 'I' that continues through time and experience, is not a mysterious thing but a very consistent, constantly updated model of our own body. The brain continuously monitors and predicts the state of our physical form — its posture, internal feelings, and potential for movement. This 'body model' acts as a basic, unchanging reference frame to which all other perceptions and actions are tied. It is the stable base on which all other learned models of the world are built, providing the needed continuity for c...

Hawkins argues that understanding is not about passively taking in information; it is about actively building predictive models. When we 'understand' an idea, a person, or a physical event, it means our internal models are accurate enough to anticipate its next state or behavior. The brain constantly predicts what it expects to sense next. Any difference from these predictions creates a 'prediction error,' which drives learning and model improvement. This continuous cycle of prediction and error correction is what drives intelligence,...

Hawkins states that current AI, despite its impressive successes in certain areas, is different from biological intelligence. Modern AI often uses large datasets and statistical relationships, lacking the brain's ability to build full, predictive models of the world using movement-based reference frames. He suggests a change in approach: instead of trying to force intelligence with bigger neural networks, AI research should focus on using the principles of cortical columns and reference frames. This means building AI systems that lear...

A major implication of Hawkins's theory is that intelligence does not have to be tied to biological brains or carbon-based life. If the neocortex uses a discoverable, universal method involving reference frames and distributed modeling, then this method could, in theory, be put into silicon or other materials. This makes intelligence less mysterious, moving it from a biological wonder to a computational event. It suggests that advanced, truly intelligent AI is not only possible but will happen once we fully understand and copy these c...

Hawkins proposes that abstract thought — understanding math, philosophy, or complex social interactions — is not a different kind of processing but an extension of the same methods used for physical perception. Just as we create reference frames for a physical cup, our brains create conceptual 'reference frames' for abstract ideas. We 'move' through these conceptual spaces, making predictions, finding relationships, and building models, much like we navigate physical space. This unified theory of cognition suggests that our ability fo...

This point highlights a key difference: the brain is a prediction machine, not just a device that reacts to input. Every moment, it forms ideas about what will happen next, based on its many internal models. Sensory input then acts as feedback, either confirming predictions or creating errors that lead to learning. This active, forward-looking nature of the brain means that much of what we see as 'reality' is actually our brain's best guess, always being refined. This predictive framework explains how we can react quickly, fill in mis...