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.
from my own experience, this is the exception, though. even in CS, if you pick some of the more theoretical problems or a not-so-hot area, you might not get funding - (almost) no matter how good you are. and that's CS. move to other subjects like sociology, linguistics, history, and you will find that barely anyone gets funded.
I haven't heard about anyone in any PhD program in US who's not funded one way or another. Sure, you might need to TA a class, but it's usually not that hard, and leaves plenty of time to do research.
