This is a slight aside, but could someone more familiar with the field than me explain what the Neural Engine is for? Does any software currently use it, or is it something that Apple are trying to push? Is it designed to build ML models, or to apply them to data?

And perhaps more basically, what for? How much machine learning is done on user machines, as opposed to renting some cloud time? What are they anticipating will be done?

I ask these questions in genuine curiousity, I assume I'm missing something but this just seems like a rather wild divergence from what (very little) I knew of the ML field.