Jeff & Matt talking about the layer data flow diagram

And I wanted to say how amazing it was to watch and how much I appreciate Numenta putting this out on the web.

The insight that mini-columns, currently used for spatial/temporal pooling in HTM, could possibly also be used for short-cutting predictive signals from layer 6 -> 4 was fantastic to watch.

PS. yeah, definitely write that down. :slight_smile:


This could be a piece of science history caught on video. The logic behind this hypothesis is very compelling, indeed.

How often, did you get to see the original discussion, in which a new foundational scientific hypothesis is born? Fantastic! Thank you for your phylosophy of openess!