Just a small note that your ability to contribute via research doesn’t go from 0 now, to 1 after you complete a PhD! As in, you can still contribute to AI Safety with research during a phd
I think you should consider more options besides PhD research and engineering.
For example, working at orgs. Or switching career paths entirely to something much more needed, such as computer security. Or joining a startup.
True, it's always good to remind oneself of a broader option space.
Could you elaborate what you mean by 'working at orgs'... since engineering would meet that definition, or do you mean explicitly other roles than engineering, such as ops or management?
I don't think I'd be a good fit for computer security, both in terms of pre-existing skills and interest, but I get your general point. (My undergrad was in physics, not CS, so I'm lacking quite a few of the traditional CS skills, except in the more theoretical subjects). Do you have a pointer to resources discussing the most needed skill sets in this broader cause area?
Apply to orgs when you apply to PhDs. If you can work at an org, do it. Otherwise, use PhD to upskill and periodically retry org applications.
You would gain skills while working at a safety org, and the learning would be more in tune with what the problems require.
I can empathise with your situation. Although I am not making career choice between engineering and research, I am looking to decide what to work on next.
I cannot give you any answer, but I can share few heuristics I use in order to decide what to work on.
For one, you are asking how should timeline affect your decision. But I would consider other variables that might affect it as well and I wouldn't place too much emphasis on only one, e.g. time in this case.
You have advantage in having binary choice here between engineering and research. So I would ask, from which it is easier to transition into the other? And which gives you more opportunities in the future?
Making decisions now based on our predictions of unpredictable future is maybe not the best way to go about it. So consider that if your predictions about future go completely wrong or in unanticipated way, which choice would be more robust?
Also if you are anything like me, deciding mostly on what you enjoy now wouldn't be very wise as what I tend to enjoy does change with time and new experiences. So I prefer choices that give me more options and easier transitions in future.
Hope this helps!
Thanks for your thoughtful answer!
In terms of flexibility, a PhD seems to score higher since it would be relatively straightforward to just stop it and do engineering instead, whereas the reverse would then require doing the full 4-5 years. I suppose it is also easier to automate an engineering job compared to research job.
Another heuristic is to choose the option where you're most likely to do exceptionally well. (Cf heavy tailed impact etc). Among other thing this, this pushes you to optimize for the timelines scenario where you can be very successful, and to do the job with the best personal fit.
Something I've been wondering while reading this is to what extent it makes sense to privilege longer timelines not based on their probability of being correct but on the expected difference you can make. I.e., you may take the pessimistic view that in a 15-year-to-AGI-world, the project of achieving alignment via the standard path is so doomed that your impact is dominated by worlds with longer timelines, even if they are less probable.
(I'm absolutely not suggesting that you shouldn't become an engineer, this really is just a thought.)
Thanks! Yeah, that's a good point mulling over. I guess it would hinge on the marginal improvement of EV in the doomed scenario to make that assertion. I don't necessarily see things being completely, hopelessly doomed in a 15-20-year-to-AGI world. But I am also uncertain as to which role is more useful in the short-timeline world, aside from an engineer being able to contribute earlier. In the medium-term timeline world it seems to me like the marginal researcher has higher EV.
So if I would be completely uncertain, i.e. 50/50, which one is better in a short timeline world, then becoming a researcher would seem like the safer choice.
TL;DR: Recent progress in AI has tentatively shortened my expected timelines of human-level machine intelligence to something like 50% in the next 15 years. Conditional on that being a sensible timeline (feel free to disagree in the comments), how should that influence my career choice?
Context:
I am currently a master student in Artificial Intelligence at the University of Amsterdam with one more year to go (mainly doing my master's thesis). So far, my go-to plan was to apply for safety-relevant PhD positions, probably either in NN generalization or RL, and then try to become a research scientist in a safety-oriented org. Given the shorter timelines, I am now considering becoming an engineer instead since that seems to require much less upskilling time, compared to doing a PhD for 4-5 years. I think the answer to my question hinges upon
I feel like I enjoy research more than pure engineering, but it's not like I don't enjoy engineering at all. Engineering seems more competitive in terms of coding skills, which I might lack compared to the most skilled other applicants. However, that is something I could practice pretty straightforwardly.
How have other people thought about this question, and how would you judge the questions about marginal value of the two roles?