Of course it doesn’t learn by itself, it has to be trained by human through automated ways mathematically (to the tune of 12 million sound scenes). But it’s well trained enough to still perform well on the other billions of sound scenarios that aren’t used to train it.
The OPN S is only better than the OPN mainly in the area of feedback prevention. So the More is better than the original OPN in the area of feedback prevention, plus it can give users more control and flexibility as to how to balance the sound scenes to the user’s preference, compared to the OPN and OPN S where this kind of flexible control is more limited.
Many people already on the OPN and OPN S only find the More improvement to be incremental, and disappointed because they didn’t get that wow factor. But I think that because of the way the More is structured where sounds can be broken down and manipulated/balanced, it’ll be easier in the future to improve this system further as opposed to the old OPN system.