I just recently learnt about the existence of this Hierarchical Temporal Memory. I already read the main document (which seems rather easy to understand), but one red flag is that the document is neither peer-reviewed nor attempting to explain why it should work in details. I tried to look around for some independent sources, and found a few papers that compare its performance against others, but none explain why it perform well or not well. I noticed some comments claiming that it was looked down by mainstream expert, but I was unable to find any actual criticisms.

So I would like to ask, what are the criticism regarding the performance of HTM? Assuming the following:

-There are huge amount of training data to use, enough for even months long training session. Basically, any criticisms regarding size or length of training is not relevant.

-Since HTM is meant to be generic, any domain-specific criticism should be related to a more fundamental problem.

Thank you for your help.