What Kind of Machine Learning Training Do You Need to Be a PhD?

A lot of those images, they answer questions about galaxy evolution and the nature of galaxies. What sort of training does it take to be capable of helping provide some labels on this data? And what sorts of scientific questions do people have about these images? I imagine they're useful for lots of different things,. But in your experience, what are some of the neat questions we can ask from this data? Why?