Here’s a simple model for doing things in life:
There’s some number of categories of things you can spend time on. (E.g. do math, give talks, go on dates, send emails.)
Whenever you spend time in any category, you’ll learn to get progressively better at doing that thing. (You’ll get more efficient, the results will be better, etc.)
After some approximately-known number of years, you’ll reach peak productivity, possibly hang around there for some time, and then your productivity will quickly decline exogenously. (Maybe you retire, maybe you’re replaced by AGI, maybe you have high altruistic discount rates.)
If you invest money, it will grow over time; and you can borrow any amount of money at the risk-free rate of return.
In this model, here are two simple heuristics for prioritisation:
When choosing what category to spend time in, you can assume that the cost and value of spending time in that category is equal to the cost and value of spending marginal time in that category at the end of your productive years, once you’ve learned all there is to learn. Because although the cost might be higher now, and the value might be lower, you’ll also speed up your entire future trajectory of learning.
When making time/money-tradeoffs, you should assume that your value of time is the same as your investment-adjusted value of time at peak productivity.
Here are three example applications of this:
If you’re thinking about giving a talk about something, and you’re very inexperienced with talks, the direct effects of giving the talk might not be worth the time-cost. But if you think that you’ll eventually do enough talks to learn how to do them quickly and well, the above heuristic suggests that you should do this talk iff it’d be worthwhile for the time-cost and impact you could achieve once you’ve practiced a ton.
If you’ll eventually become really good at doing something hard (like doing research in some field) you should focus most of your time on it, and pass up on doing tasks that let you have more impact now, but where you won’t improve as much in the long-term (either because you won’t spend enough time on it or because there’s not much room to grow).
If you think your salary will increase much faster (from learning) than the investment rate of return, you should be happy to buy time at a much higher rate than your current salary. (Mark Xu makes this point here.)
And here are some big caveats:
For the money point, you might be liquidity constrained from not being able to borrow at the risk-free rate of return, or you might be risk-averse and not confident that you’ll reach your later higher salaries.
When choosing what to spend your time on, a big fraction of value is information value, which would shift around what things you’ll eventually become really good at. The above heuristic is kind of silent on this.
For many types of tasks, the things you should do to get good are quite different from the things you should do to exploit being good. For example, for giving talks, you might actively seek out low-stakes opportunities to practice early on (e.g. toastmaster sessions), but seek out high-stakes opportunities once you’re good. So even if the above heuristic recommends that you should give talks, make sure to not trust it about which talks to give.
I think the altruistic discount rate on EA work is actually really high. This is mainly because there will be significantly more EAs in the future, and there are some things that only current EAs can work on, like (i) short AI timelines, (ii) early community building, and (iii) starting work that requires a lot of serial time. So taking into account both learning (making you more productive later) and this discount rate (making you more productive sooner), your peak altruistic productivity might be soon (or now!). In that case, this heuristic differs much less from just doing what seems best in the short-run.
Also, a note on attitude: This heuristic will sometimes recommend that you prioritise as if some tasks are easy and high-impact, even if right now, they’re in fact really hard. If you do follow this recommendation, remember that the reason that the task is nevertheless worthwhile is that learning and shifting forward all your future work is super high-impact. So acknowledge and celebrate your hard, super high-impact work accordingly!