Posts
Comments
As someone who gives data science interviews, my (personal, unreliable) opinion is that you should start preparing for interviews as soon as possible, and actually begin interviewing as soon as you feel ready.
I'm not saying you'll get in on the first try! You might, in which case you'll save a lot of effort doing anything else. If not, you'll get some sense of what the interview process is like, and where your strengths and weaknesses are.
If you can't get interviews at all, you may need to think about improving your resume. That could look like options 1, or 2, or 4 if you can swing it; the details probably depend a lot on your personal circumstances.
If you can get interviews, but not jobs, you should probably work on your interview technique. For early-career hires, we care more about how the interview and practical exercise go than anything else. (Remember to ask the interviewers for feedback at the end, e.g. "is there anything you think I could improve on?")
If you want to go get some super-impressive experience, that's not a bad thing, it's certainly going to make us more interested; that said, it's a large amount of work to do so convincingly, and it won't save you if you can't impress on the more routine parts of the interview.
Also, don't feel you have to sell your existing experience short: "I did some clever feature engineering that resulted in a better model for our data" is actually a pretty good answer. I can't speak for AI safety, but there are lots of other opportunities that would be happy to have someone who knows their way around a dataset.
If you're not sure how to explain it, then practise that! You're going to be evaluated on your communication as much as anything else, and explaining technical concepts to people who don't understand them is often part of the job. They won't need to know it inside-out, just give them a sense of what's going on, and why your efforts mattered.
Well, the motte is "I'm very epistemically humble", and the bailey is "that's why I'm always right".
No, not really? I generally ignore anything Scott writes which could be described as 'agreeing with Yud' -- it's his other work I find valuable, work I wouldn't expect Yud to write in any style.
I think there's a tendency to assume the rationalist community has all the answers (e.g. The Correct Contrarian Cluster), which seems (a) wrong to me on the object-level, but also (b) at odds with a lot of other rationalist ideas.
If you point this out, you might hear someone say they're "only an aspiring rationalist", or "that's in the sequences", or "rationalists already believe that". Which can seem like a Motte and Bailey, if it doesn't actually dent their self-confidence at all.
I'd like to know that too! I've had some fairly moving meditative experiences, but still find it oddly aversive to do; strength training feels easier, just because my reward system seems to understand it better.
I think joining a meditation class can help, as you get a social context and a schedule, but that does depend on having a good teacher nearby.
Maybe strength training? It made a big difference for me, and I wouldn't have expected that going in.
The key is to do it regularly, even if you don't do a lot; ten minutes every other day is fine, and you can do it at home with bodyweight or resistance cables.
(Meditation is also great, but I assume most LWers will have thought of that.)