Generalization -- Fast and Slow (incomplete)

I don't think I ever fleshed out this idea. I was trying to point out the multi-frequency and continual nature of ML optimization and that human development should be considered part of the optimization process just as machine learning is.