finding the right problem

  1. Decide on a problem. There are a lot of problems in the world. Which one you choose really matters. If you’ve got no clue which problem you should be prioritizing, read this list from 80,000 Hours. If no problem immediately strikes you as the most important, spend a few days/weeks/months researching your top choices. Consider using a prioritization strategy like the ITN framework, which ranks problems based on their Importance, Tractability, and Neglectedness.
    • “Months” might seem like a long time, but actually, even if you spent 3 months researching which problem to solve, you’d only be spending 0.625% of the average 40-year career doing planning (notwithstanding this whole AI thing, which might mean you only have 3? but anyway…).
    • Say you have 100 minutes to work on a project. If you think spending 5 minutes of that time planning out what specific steps you’ll be taking before you start is reasonable, then “months” makes total sense to plan out what is probably going to become your most important contribution to the world: your career.
  2. After you’ve decided what problem you want to focus on, stare at the problem. Figure out what the cruxes are, where intervening would be maximally effective. Ask yourself: why do I think [insert thing— e.g. AI safety] is going to go badly by default? Why is this the case? What needs to change for me to update my belief on this?
    • Write up a document with all of your key assumptions and the gaps you’ve identified, and then send it to people working in the field to ask for feedback/stress-testing.
  3. Choose one (or multiple) of the cruxes you’ve identified (e.g. in AI safety, you might choose something like “AI will enable authoritarian power concentration” or “AI will lower barriers to entry for CBRN weapons development”) and brainstorm projects that could help solve the problem.
    • I recommend doing a project brainstorm before searching for orgs/people who are working on solving the problem; if you can come up with a sufficiently promising novel project, (1) you can probably find a funder and (2) your counterfactual impact will be higher than if you just applied to join an existing project, since you’d be filling a gap that no one else was working on before you.
  4. Shortcut: If you want to skip thinking about problems and how to solve them, consider reading RFPs from grantmakers! They’ve thought about problems and what needs to be done to solve them, and as an added bonus, you can apply to them directly for funding to do the projects they directly endorse doing.
    • However, I’d still recommend doing the deciding and staring etc. if you have the time; it’s a useful exercise in reasoning from first principles and can help you determine what your values are/what you care about based on the problems you’re drawn to. Do you care about x-risk? S-risk? How expansive is your moral circle? Are you more drawn to problems in your local community, or do you want to work on something that matters at a civilizational scale?

being agentic about solving the problem

  1. You are probably not fully considering your option set at the moment. There are many unknown unknowns. For example, I spent quite a long time waffling between two career options (startups vs. policy), trying to decide which one would be more impactful given I want to contribute to AI safety. I didn’t even consider donor advisory as something in my option set, but it is! (And now I’m waffling between three career options.)
    • Useful exercises here:
      • Set aside 5-10 minutes to brainstorm a list of every possible thing you could be doing with your time. Err on the side of inclusion, even if you think you don’t have the resources to do a thing or it is too ambitious. Listing out your options will train you to think more expansively and intentionally about your option set.
      • Surround yourself with highly agentic, motivated, and ambitious people who are trying to achieve the same goal as you. Observe what they do, and sometimes you’ll think to yourself, wow, I would never have thought to do XYZ! Then you’ll know to consider it for the future.
      • List out all of the resources you have available to you, and then feed that list into an AI, tell it about your goals, and ask it to list all of the options you might want to consider to achieve said goals.
    • Heuristic to internalize: Everything is in your option set because you can send cold emails.
  2. Ask for help. Talk to people who have way more resources than you. Talk to people who are in the field you want to enter, or in the org you want to apply to work at. Ask them for career advice. Ask them what they think the most highly impactful thing you could do is, and then, if you end up doing the thing, follow up with them and ask for help!
    • People like to help people who want to do good.
  3. By far, the most important thing to do is practice. Just go out into the world and try to do the thing!
    • Heuristic: If you haven’t failed at doing something yet, you’re probably not at the upper bound of what you could be doing. Court rejection, and see it as a data point for calibration. Testing the limits here is the only way you can determine where they are.

Now stop reading this post, and go do the thing.