Thousand Brains — cortical reference

Hawkins' thesis is simple and unsettling: every cortical column learns a complete model of the world, and what we call perception is a vote among thousands of them.

There is no "center" where the object arrives finished. There are many partial models, each with its own spatial reference frame, converging.

Why it matters for AI

Current LLMs are largely monolithic: one big model producing one next prediction. Hawkins' architecture suggests a fundamentally different path — distributed voting among models with their own spatial references — may be necessary for behavior we recognize as cognition.