Pick one (1) technical subject[1]. Read the textbook carefully (maybe take notes). Do all the exercises (or at least try to spend >20 minutes on exercises you can’t solve). Potentially make flashcards. Study those flashcards. Do the real thing.[2]
I regret having spent so much time reading philosophy, and not learning technical subjects. I have gained remarkably little from “learning how to learn” (except the stuff above) or productivity or epistemology (excluding forecasting)[3]. I remember reading about a heuristic (might’ve been on Gwerns site, but I can’t find it right now): Spend 90% of your time on object-level stuff, 9% of time on meta stuff, 0.9% of time on meta-meta stuff, and so on).
Learning forecasting is great. Best learned by doing a thousand forecasts (flows through to probability theory).
I think linear algebra, causal inference or artificial intelligence are good candidates. I am unsure about game theory, it’s been useful only in metaphors in my own life—too brittle and dependent on initial conditions. But in general anything where you can do exercises (so most things from 6.) and have them be wrong or right is good (so stuff like coding is better than math because checking a proof depends on knowing what a good proof looks like).
Some advice (with less justification):
Pick one (1) technical subject[1]. Read the textbook carefully (maybe take notes). Do all the exercises (or at least try to spend >20 minutes on exercises you can’t solve). Potentially make flashcards. Study those flashcards. Do the real thing.[2]
I regret having spent so much time reading philosophy, and not learning technical subjects. I have gained remarkably little from “learning how to learn” (except the stuff above) or productivity or epistemology (excluding forecasting)[3]. I remember reading about a heuristic (might’ve been on Gwerns site, but I can’t find it right now): Spend 90% of your time on object-level stuff, 9% of time on meta stuff, 0.9% of time on meta-meta stuff, and so on).
Learning forecasting is great. Best learned by doing a thousand forecasts (flows through to probability theory).
I think linear algebra, causal inference or artificial intelligence are good candidates. I am unsure about game theory, it’s been useful only in metaphors in my own life—too brittle and dependent on initial conditions. But in general anything where you can do exercises (so most things from 6.) and have them be wrong or right is good (so stuff like coding is better than math because checking a proof depends on knowing what a good proof looks like).
I predict you won’t finish the textbook. No problem.
I think I learned more from a course on social choice theory than all philosophy from before 1950 I have read.