Jeff Hawkins Mini-Columns Hypothesis?

I found 3 talks from 2020 on your tube on a hypothetical idea of mini columns that Jeff Hawkins was working on. Here’s a screen shot a slide form two of the videos:

I’m wondering if this is something Jeff is still working on and if so is there any other talks or papers out there for me to read?

Thanks for bringing these back to the fore. I had forgotten about them and its good to be reminded of the theoretical work.

They haven’t done a lot of recorded discussions for a couple years so its not clear where they are on the theory. They’ve also devoted a lot of effort to applying their tech to Deep Learning so there might be a lot less theory than there used to be.

Re-reading these slides, I struck by how much I hate Numenta’s use of the word “spatial” since its never clear what they mean and it kind of conflicts with my intuition of Euclidean space.


It’s not just Numenta, there is a rampant abuse of “dimensions” in ML. And no one cares about a functional theory,

1 Like

what d’you find confusing about dimensions in DL?