Disclosure: first-year CS Ph.D. student. I would say, no, a scientist doesn't really need a "Ph.D.", but does need the training that a Ph.D. involves. To me,

Disclosure: first-year CS Ph.D. student. I would say, no, a scientist doesn't really need a "Ph.D.", but does need the training that a Ph.D. involves. To me, this question is very different from the question of whether someone who wants to be an industry software engineer needs a bachelor's degree (a popular HN topic). The difference is that in a science Ph.D. program, you are doing science the entire time you are there. Sure, you take a few classes as well, but these are only to reinforce concepts you'll need in your research; the entire reason you're there is to learn to do scientific research by actually doing scientific research. You are doing at school what you'll be doing in your career; you're just getting paid less for it because you're "in training". I don't see any reason why one couldn't get a similar education by doing scientific research in an industry which does not result in a formal "Ph.D." degree, rather than an academic setting that does result in a Ph.D., it just doesn't seem like it's been tried up until (apparently) now. Personally (all other things being equal) I'd rather do a Ph.D. program in an academic setting because an industry setting would likely be constrained to research areas that are beneficial to the company.