“In a pure hierarchical model, these connections would not be necessary. In fact, more than 95% of the synapses in the neocortex are not explained by the pure hierarchical model. Additionally, many AI and deep learning networks are built on this hierarchical model. They typically require dozens of levels and millions of examples to learn something, whereas a human can learn something new in just a few exposures. The brain is clearly doing something different, and the general hierarchical view explains some of the story, but not all, so there must be something missing.

We believe we have uncovered the missing piece: all cortical columns have a signal representing location.”