Hawkins’s Thousand-Brain Remembered Modeling

Technologist/neuroscientist Jeff Hawkins’s Thousand Brains Theory is primarily a theory of neocortical intelligence, not a full metaphysical theory of phenomenal consciousness; its relevance to consciousness lies in explaining how massively parallel cortical models yield the unified, stable, object-centered perceptual world we experience. Hawkins proposes that the neocortex does not build one central model of reality, but thousands of concurrent models distributed across cortical columns, each using sensorimotor reference frames to model objects, places, concepts, and relations (Hawkins, 2021; Hawkins et al., 2019). Consciousness, on this view, is a brain-generated phenomenon dependent on neocortical modeling, voting, prediction, and action, though Hawkins does not claim to solve the hard problem of why such modeling is accompanied by felt experience.

Cortical Columns as Modeling Systems

Hawkins’s central anatomical-functional claim is that the human neocortex comprises roughly 150,000 cortical columns, each independently learning complete models of such objects and concepts using map-like reference frames, with unified perception arising not from any central processor but from a consensus the columns reach by voting. Rather than assigning perception to a single hierarchical representation, the theory distributes model-building throughout the cortex, so that “every part of the neocortex learns complete models of objects and concepts,” not merely local features (Hawkins et al., 2019). On this Thousand Brains Theory, what we perceive moment to moment is the brain's continuously updated simulation of the world rather than the world itself, since nothing reaches the cortex but electrical spikes.

Consciousness, for Hawkins, decomposes into two tractable components: awareness, which he identifies with the continuous formation and recall of memories of the brain's own recent model states, and qualia, which he traces to what structurally uniform cortical tissue happens to be connected to. He accordingly predicts that any system that learns a world model and continuously remembers and recalls its states will be conscious, and that the hard problem will dissolve rather than be solved (Hawkins, 2021).

Columns, Reference Frames, Voting

Building on Vernon Mountcastle's proposal that the neocortex executes one common algorithm throughout (Mountcastle, 1978), Hawkins argues that what differentiates cortical regions is not their structure but their connections — the circuitry looks remarkably alike across visual, language, and touch regions. Each column, he proposes, learns complete models by attaching reference frames to objects, adapting the grid-cell and place-cell machinery of the entorhinal cortex and hippocampus to the cortex at large (Hawkins et al., 2019). Because reference frames require both sensory input and motor output, knowledge is inherently sensorimotor: modeling requires movement. Columns are richly interconnected and integrate their outputs, so that thousands of partial, parallel models converge by voting into a single perceptual result (Hawkins, Ahmad, & Cui, 2017).

Perception as Simulation

A direct consequence concerns the status of experience. Cortical input is a distorted and incomplete patchwork, yet perception is uniform and complete; Hawkins takes the voting mechanism to explain this singular, non-distorted perception, offering a mechanistic account of perceptual unity that addresses what others treat as the binding problem. Most cortical prediction is unconscious: when predictions are met, nothing is noticed, and only mismatch draws attention and prompts model revision. What one experiences is therefore a simulation generated by the brain's model—a model that includes a model of the self.

Awareness as Remembered States

Hawkins's positive account of awareness identifies it with memory. His illustrative case is an action performed and then erased from memory: on discovering a video of oneself doing it, one would judge that one had not been conscious during it. Moment-to-moment awareness, on this view, consists in a continuous sequence of memories of the brain's recent states, recalled and driving ongoing action. The felt sense of being conscious is thus a function of the cortex's continuous self-recording, and Hawkins generalizes this into a criterion: a system that models the world, continuously stores the states of that model, and recalls them will be conscious.

Qualia and The Deflation of The Hard Problem

Hawkins explicitly acknowledges the hard problem (Chalmers, 1995) and devotes a chapter to it, but treats it as a temporary rather than principled obstacle. His approach to qualia follows from cortical uniformity: since the same circuitry underlies vision, touch, and hearing, the qualitative differences among modalities reduce to differences in what a column is wired to—which inputs it receives and which behaviors it drives. He anticipates that attitudes toward consciousness will shift as such accounts mature, so that while remaining questions persist, the hard problem will cease to be regarded as a problem at all. Reviewers sympathetic to the neuroscience have nonetheless found this treatment unsatisfying, objecting that identifying qualia with connectivity patterns restates the correlation rather than explaining why any such pattern is felt.

Machine Consciousness, and Standing Objections

Because consciousness on this account requires only the right functional architecture, Hawkins holds that machines built on Thousand Brains principles would be conscious, while remaining affectless—lacking the evolved drives of the older subcortical brain—and argues that unplugging such a conscious machine would raise no concern. Two objections recur. The philosophical one is that his account addresses access, unity, and self-modeling while leaving phenomenal character unexplained, a gap his own framing concedes but declines to treat as foundational. The neuroscientific one targets his sharp old-brain/new-brain division, which critics link to the discredited triune-brain schema, noting that the mammalian neocortex is closely related to the pallium of birds and reptiles, with comparable circuitry and no evident intelligence deficit at matched neuron counts.

References

Chalmers, D. J. (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200–219.

Hawkins, J. (2021). A Thousand Brains: A New Theory of Intelligence. Basic Books.

Hawkins, J., & Ahmad, S. (2016). Why neurons have thousands of synapses, a theory of sequence memory in neocortex. Frontiers in Neural Circuits, 10, 23.

Hawkins, J., Ahmad, S., & Cui, Y. (2017). A theory of how columns in the neocortex enable learning the structure of the world. Frontiers in Neural Circuits, 11, 81.

Hawkins, J., & Blakeslee, S. (2004). On Intelligence. Times Books.

Hawkins, J., Lewis, M., Klukas, M., Purdy, S., & Ahmad, S. (2019). A framework for intelligence and cortical function based on grid cells in the neocortex. Frontiers in Neural Circuits, 12, 121.

Mountcastle, V. B. (1978). An organizing principle for cerebral function: The unit module and the distributed system. In G. M. Edelman & V. B. Mountcastle, The Mindful Brain (pp. 7–50). MIT Press.