[ Question ]

Career Advice?

by sty.silver 1 min read6th Jan 20203 comments

6


I'm an aspiring rationalist / effective altruist / mathematician, on track to finish a master in computer science in a couple of months. I'm trying to figure out what to do next; that's the decision.

Various considerations, primarily reading articles on 80000 hours have led me to decide that I want to do a PhD* in Machine Learning, with the goal of working on technical AI safety, or earning-to-give as a backup plan. With that decision already made, I think I don't want to apply for 1-on-1 counseling (obviously I would take the spot off of someone else), but I'm still left with a couple of questions that I don't really know how to approach.

(1) * This is a relatively minor point, but – I live in Germany, and the more common degree here is something called a doctorate. Supposedly it is equivalent to a PhD but the process differs a bit in that the doctorate is even less structured and even more focused on individual research. I've read that many universities do offer the real PhD though. Ignoring the difference in the process, I'm wondering whether I should or shouldn't worry about which degree will be more valuable, and maybe try to find a PhD program rather than the regular doctorate.

(2) According to this site, my university ranks #126–150 in computer science and #251-#300 worldwide. This is not super high – my question is, how much does this matter? Has anyone credibly tried to quantify this? Also, are these rankings even reliable, or should I use another site?

The two options I'm primarily considering is staying where I am or moving to Berlin. Berlin's highest-ranked university in computer science is at #55.

(3) Related to (2), I don't really know how difficult it will be to actually get a PhD / doctorate position. 80000 hours has some resources on this, but they generally assume that the goal is to get into one of the top 5 universities worldwide, and in fact they mention at one point that anything below that might be a lot easier. Obviously #55 is not anywhere near the top 5, so basically I'm wondering whether just finishing a masters with good grades is realistically enough, or if one really does need to have published stuff already. The university website doesn't answer this; they just tell you the formal requirements. The answer here might just be "it depends", but I don't actually know that.

(4) I don't like the machine learning institute at my university; I have issues with their lectures and particularly the lack of rigor. This is one of the reasons I'm considering moving. On the other hand, another professor (whom I do like) has more or less offered me a doctorate position that would be about something like "privacy-preserving machine learning", i.e. a hybrid between ML and information security.

I'm probably not going to take this because I imagine it'll be a lot less valuable than a pure ML PhD (?), but I'm not sure yet. It would be awfully convenient.

This mostly summarizes where I am right now – any input is highly appreciated.

Meta: I was going to post this in an open thread rather than make it its own thing, but there don't seem to be open threads anymore. I think this definitely raises the bar for asking things – if one has to start a new topic, they might worry about whether it's appropriate and such, and end up not doing anything instead.

6

New Answer
Ask Related Question
New Comment

1 Answers

Their career review for ML PhDs says

Want to use an ML PhD to make the world a better place? We want to help.
We’ve coached dozens of people considering PhDs, and can often put you in touch with relevant experts for more guidance. Apply for our free coaching service, particularly if you want to work on AI safety

I'd say go ahead and apply. They can decide whether they want to speak to you (I'd guess that they would). If they don't, no big loss, and if they do it only takes about about 45 minutes of 1-1 time.