On intelligence

About the book

On Intelligence sets out a theory of how the neocortex works and what that theory implies for building intelligent machines. Jeff Hawkins, writing with science journalist Sandra Blakeslee, draws on his engineering background and on neuroscience to argue that prediction is the cortex's central job. The book moves from anatomy and physiology toward a general framework he calls memory-prediction.

Hawkins proposes that the cortex stores sequences of patterns and uses them to anticipate incoming sensory input. Perception, on this account, is largely a matter of matching what arrives against stored expectations and noticing the difference. He extends the idea across vision, hearing, touch and language, treating them as variations on one mechanism. The later sections turn to artificial intelligence, arguing that systems modeled on this predictive architecture would behave more like people than systems built around logic or hand-coded rules.

The neuroscience is presented for a general reader, with diagrams and plain analogies rather than equations. Hawkins builds the prediction argument step by step, so skipping the anatomical groundwork makes the later proposals feel arbitrary.

How to read it

Watch for where Hawkins moves from describing known brain structure to proposing a model. The two are not always clearly separated, and the shift is where the argument asks for the most trust.

The prose is direct and example-driven, closer to an extended essay than a textbook. Expect repetition of the core idea from several angles, which helps retention but can feel slow. Treat the AI chapters as a historical argument rather than a current technical survey, and note which predictions have aged well and which have not.

Who it suits

This suits readers curious about the biology behind machine intelligence, especially those who want a conceptual framework rather than code or mathematics. Engineers and product-minded technologists will find the memory-prediction idea useful for thinking about perception and sequence. Students of neuroscience get a readable statement of one influential hypothesis, and general readers who enjoyed popular brain science will find the pace manageable.

Reading considerations

The book argues for one theory rather than surveying the field, and specialists have debated how far the memory-prediction framework explains cortical function. Its framing of artificial intelligence predates much of the current landscape, so read the technology claims with that gap in mind.