An important caveat I think: your mentor is critical. I’m a PhD student, and I have an incredibly lucky mix of subject matter and mentorship (my mentor has actually gotten some awards in the past for being a good one). (Also, he’s funny.) Every day my research is a blast. My mentor is truly an expert in his field (he’s actually mentioned in Norvig’s AI textbook introduction) and has the answers to my questions. I get just enough steering to make sure I’m on track to producing something novel, but I’m totally free to explore. This is all bolstered by me being truly extremely intrinsically interested in my sub field. It’s really a very pleasant experience. I’m not terribly worried about the prospects afterwards as furthering humanity’s knowledge is rewarding enough to me. Heading to industry as a code monkey afterwards wouldn't upset me as I’m really just having fun as a PhD student, and it’s all because my mentor is incredible.

It’s important to try working with multiple mentors if you can. I actually started with a different person, and he probably would have been considered by most to be ideal. I was basically totally funded to do whatever I wanted with very little mentor interaction. Interestingly, this wasn’t great for me. In part I wasn’t interested in the subject, but more than that I greatly benefit from weekly mentor interaction to check in with my direction. Now I don’t have funding (working to support myself while doing research), but I’m happy as a pig in mud.