3 ms·
> But when you want to do research, design great composible systems, be creative, write good code, it is better to do it casually. As a fellow ML researcher, I
by godelski 3y ago
> But when you want to do research, design great composible systems, be creative, write good code, it is better to do it casually.
As a fellow ML researcher, I actually have a good answer to this. It's a creative task, not a labor task. Creative tasks require your brain to be exceptionally flexible and to view things from unique perspectives. Research is literally creating something that no one has thought of before so you clearly have to do things no one did before. Plus, the truth is that if you're a researcher, do you ever really "stop"? Because I'll even be sitting down playing videogames and sometimes thats when ideas come to be. Other times when I'm out for a walk. Sometimes it's also when I'm deep in the codebase tearing stuff apart because it's already 5 levels of spagetti figuring out exactly what each line does and thinking a lot about it and its connection to the rest of the program (seriously, we could do a lot more testing if we'd rewrite code from the ground up. Not doing this is like thinking your first attempt at anything is good. Idk about you, but my first attempts are absolute shit and I've learned a lot by the next go around). But all those things are connected together. Because science is very much an art. My best ideas always come from after time I've taken off which is usually long overdue (as in I can barely name 2 realistic vacations in the last 5 years), but exactly like you say, it feels like I can't take the time.
> The moment I try to talk about the creative type of work my colleagues look at me like I'm making excuses or that I'm a slacker.
Just me and the teams I've worked on/with or are these people also typically low productivity and not the most creative? The type of people that treat science as something you can just hammer away at like there's a clear algorithm to solving any problem. I know it's wrong, but that shit gives me mad imposter syndrome.
Like our literally job is about optimization. You think we'd be able to at least optimize a work day, or maybe a month?
- JonChesterfield 3y ago> Like our literally job is about optimization. You think we'd be able to at least optimize a work day, or maybe a month? There is lots written about this. Whole careers built on advice about it. It seems what works for some people does not work for others. My outline won't help you if you're in the valley but I'll write it out anyway, some people live elsewhere. 1/ Wake up early. Like daybreak early. 2/ Do not look at email, slack etc. No admin stuff at this point. 3/ Leisurely breakfast letting your mind wander. Historically cooked by wife, ymmv there. 4/ Spend the morning walking, thinking, writing things down. Especially write ideas down before lunch. 5/ It's midday. Optionally eat things. Open email. 6/ Entire afternoon is spent doing whatever "work" one found in the email. Put out the fires. Do the code reviews. Write whatever simple stuff seems to be required. 7/ It's evening. Try to forget the fires. Exercise helps. 8/ Spend the evening on research or socialising. Aim is to have interesting context for dreams. 9/ Sleep time. That is relatively new to me - couple of months or so - and working really well. Note the complete lack of family activities - either that overlaps the afternoon fires or the morning creativity, or both. Opportunity cost is rough. No creative work for you if there's a toddler or an unsupportive partner present. This morning-creative afternoon-productive split works really well if most of your colleagues are five hours west of you. They wake up in time to light things on fire. If they're five hours east, maybe swap the two blocks of time over. If they're in the same timezone and expect to talk to you all day, no creative work for you. Opportunity cost again.