Sharing an analogy I consider interesting

[Bitking]

Surprise and free energy minimisation?
You may want to read this:

Does predictive coding have a future?

Clearly - in this forum, we take the answer in the affirmative! I think this Paper speaks directly to the theoretical underpinnings of HTM theory. Fristen talks of the difference between predictions and perceptions (I call it surprise) as the error signal that is minimized by training. You learn surprising things - exactly what an anomaly is. Some people have expressed some reserve as the underpinnings of HTM can’t be quantified. Clearly, the variational free energy is measurable & calculable…

There has been a lot of work in this area.