That's why I said you have to find a lab that does what you're interested in. I have done just that, and for the last 5 years as a phd student I've been paid mostly to do what I would do as a hobby otherwise (read/implement deep learning papers), and try to advance state of the art.