The HTM theory is being developed based on a few assumptions. As far as I understood, one of these is that the neocortex is the brain structure responsible for “intelligence”. The consequence is that it should be enough to study the neocortex to develop the HTM theory. Another assumption is that the neocortex seems to be quite uniform. As a consequence, it should be enough to study one cortical column to understand how (human) intelligence works.
What are other assumptions that the HTM theory makes?
One that seems obvious is that HTM theory assumes that all the cells in each layer are doing the same predictive cell thing.
It seems to me that there are at least three population (and likely more) that I see doing different things:
the classic HTM predictive neurons in the feed-forward direction.
the deepleabra predictive neurons in the feed-back direction. These work very closely with The thalamus.
the grid-forming / sparsity neurons in layer 2/3.
I see all these working in a synergistic way.
I also think it is very odd that the HTM models and diagrams seem to ignore the huge tracts to and from the sub-cortical structures. These are very important to the deepleabra model to implement hierarchy, and I suspect, to the HTM predictive model as a synchronizing element.