In my experience, the PhD does makes you smarter in a sort of trial by fire. I haven't met a single (math/CS) PhD student who wasn't intimidated by the magnitud

In my experience, the PhD does makes you smarter in a sort of trial by fire. I haven't met a single (math/CS) PhD student who wasn't intimidated by the magnitude of their research field when they started, or seriously intimidated by the math in a seminal research paper in their area, but came out with a mastery of the subject (which is not to say that there aren't any crap PhDs -- there are). At the very, very least, if you finish your PhD diligently, or even get close to finishing it, you will develop a confidence that you can in fact chip away at the most abstract and difficult problems, and usually come up with a good solution, while building on the efforts of much smarter people. At a very practical CS/algorithmic level, a lot more will seem possible, and you'll reinvent the wheel a lot less. I've often seen very smart people in CS/ML who didn't have a PhD. However, they sometimes also lack the body of knowledge in advanced algorithms or statistics that comes from 5-6 years of monk-like reading and chipping away at immense intellectual problems that have been tackled by some of the smartest people in the world. So if you don't mind 5-6 years of near-poverty, intensely difficult problems, a possible lack of interest from your advisor (thankfully not in my case), peer-review rejections that sometimes seem arbitrary and disheartening, and can take 3 months to come around each time, large minds and frequently larger egos, intense orgasmic breakthrough eureka moments that come from spending months or years working on a problem, self-learning some truly amazing things, and drinking discount "St. Remy" brandy instead of Remy Martin, then by all means go in for a PhD -- you'll come out sharper if you make it.