Summary of On Intelligence by Jeff Hawkins

Jeff Hawkins' On Intelligence (2004) presents a revolutionary theory of human intelligence centered on the neocortex as a memory-prediction system. Rather than viewing intelligence as computational power, Hawkins argues that true intelligence emerges from the brain's ability to store experiences and predict future events.

Core Theory: Memory-Prediction Framework

Intelligence arises from the brain's constant prediction of future inputs based on stored memories
The neocortex operates as a unified, hierarchical system using the same algorithm across all regions
Information flows both up (sensory input) and down (predictions) through cortical layers

Key Brain Principles

Hierarchical Memory: Lower levels process simple patterns while higher levels handle complex, abstract concepts
Invariant Representations: The brain extracts essential features, allowing recognition despite variations in appearance
Temporal Sequences: Memory stores experiences as ordered sequences, not static snapshots
Prediction and Learning: When predictions fail, the brain updates its model of the world

Implications for AI

Traditional computers lack the brain's learning and contextual capabilities
Current AI approaches miss crucial elements: feedback loops, temporal processing, and hierarchical memory
True machine intelligence requires mimicking neocortical principles rather than increasing processing power

Hawkins envisions intelligent machines built on brain-like architecture that would think without human emotions or survival drives, making them powerful tools rather than threats.