A Thousand Brains

Hawkins argues the neocortex learns through many small models, each tied to its own reference frame.   He then draws conclusions for how AI should be built.