Summary

Choose a career by looking outward: get good at something that helps others. A dream job combines engaging work, meaning, mastery, supportive colleagues, and basic conditions. To maximize impact, work on problems that are large in scale, neglected, and solvable, then contribute through research, advocacy, government, organization-building, or earning to give.

Treat career choice as an iterative decision process, not a one-time act of self-discovery. Build career capital – skills, connections, credentials, character, and runway – while testing personal fit through cheap experiments. Explore high-upside paths early, develop skills that are hard to automate and complementary to AI, and review at a regular cadence.

“Plans are useless but planning is essential.”

“On your deathbed, helping people is one of the things you’re least likely to regret.”

“If we really want to be satisfied, we have to stop focusing on ourselves and our own passions and instead turn our attention outwards to what we can do for other people.”

Chappy’s Review

My new go-to for stepping back to think about my career and its impact! Clearly lots of good thinking that went into this and love the frameworks + tactical guides.

Key Takeaways

• Impact = number of people helped × magnitude of improvement × duration

Notes

Introduction

• Study: the best charities are 100x more cost effective
• Worth investing 800 hours intro a 1% improvement to career trajectory
• Follow your passion < master valuable skills + use to do meaningful work

1: What to Aim For

1: What Makes for a Dream Job

• 3 ways to”follow your passion” is misleading:
  1. Passion not relevant to career
  2. Need more than just passion
  3. Needlessly limits options
• Better to find passion in a chosen pursuit than vice versa
• We’re not great at predicting what’ll make us happy
  • Eg. peak-end rule
• Mainstream career advice:
  1. Un-stressful
  2. Work-life balance
  3. Job security
  4. Highly paid
• ↑ resulting top of the list: insurance actuaries 😂
• Money + stress potentially counterproductive
• Optimize for flow (challenges you)
• 5 criteria for dream job (day-by-day):
  • Engaging: day-to-day enjoyment, flow / challenge, autonomy
  • Meaningful
  • Good at: skill / mastery
  • Supportive colleagues
  • Meets basis needs (commute, hours, pay, job security)
• Slogan: “get good at something that helps others” / “do what matters”

2: How to Make a Difference

2: Quantifying career impact

• Doctors not effective
  • Eg. >150/100k doesn’t impact DALYs, more just speculative procedures
  • Estimate: 3 lives saved
• Positive impact definition:
  1. Number of people whose life your improve
  2. How much you improve them
  3. How long the effects last
• What does it mean to lead a moral life:
  1. Have good character
  2. Do what’s right
  3. Make the world better (↑)
• Luck, eg. Petrov Cold War nuke saving billions of lives

3: Impact outside career

• Eg. volunteers costs training staff, but can perpetuate its own incompetence
• 3 options:
  1. Donate effectively (eg. UBI, global health, education)
  2. Voting / advocacy
  3. Be a multiplier to help others be more effective
• Eg. 70k is the top 1% globally
• Eg. community organizing
• Eg. helping an expert researcher
• Donating to an environmental charity is 70x more effective than reducing own carbon footprint

4: How to Pick a Problem

1. Is this problem large in scale?
2. Likely to be neglected?
   • Low-visibility, low-probability
   • Least protected by current system
3. Is progress possible + proven?
   • Even if “hit-based” / lumpy

• Ask “Why aren’t people already working on it?”
• Eg. switching to global health
• Eg. non-human lives?
• 5: What’s the World’s Most Pressing Problem?

5: World’s Most Pressing Problems

• Longterm-ism: the view that helping future generations should be a key priority of our time
• Even the very worst case environment change won’t be existential
• Biosecurity is pressing
• AGI
  • Intelligence + technological + industrial explosions
• AGI risk - loss of control:
  1. Goal specification
  2. Instrumental convergence
  3. Reward hacking
  4. Deceptive alignment
• Expert guesses: 5-25%
• AI safety still has <1k researchers, so great early vet
• Other ideas:
  • Gradual disempowerment: systemic forces - AI converting to economic output - humans in loop squeezed out
  • AI-enabled concentration of power: ↑ but for whole countries + societies
  • AI legal rights
  • Regulating weapons of mass destruction
  • Regulating colonization of space
  • Building AI to help us solve ↑ ↓
  • Global priorities research: finding + refining ↑

6: Which Jobs Help People the Most

1. Earning to give
   • Eg. Google SWE can pay for 2 non-profit ED salaries
   • However, morals first
2. Communicate ideas (advocate)
   • Inherently more solo
   • Strong power law
   • Can also be part-time
3. Research
4. Federal employee
   • Eg. responsible for large budget
5. Building organizations
   • Talent > funding

• Beware of risks, eg. moral backsliding

3: Career Path

7: Career Self-Sabotage

• Build career capital first
• Income + impact usually peaks in 40s-50s
  • Even in most experimental fields its age 30
• Career capital:
  • Skills
  • Connections
  • Credentials
  • Character
  • Runway
• What’ll get you there fastest?
  • Ask people a couple steps ahead of you
• But don’t wait too long to have an impact
• Prioritize:
  1. Talent match
  2. Learning environment (eg. mentors, feedback)
  3. Time spent practicing
  4. Right place right time
• “You’re one of the most talented people I know. Where do you hang out?”

8: Best Skills for the future

• 2 ways automation doesn’t decrease jobs or wages:
  1. Increases value of other skills
  2. Increases productivity, hence value of employee
• Depends on if truly all jobs can be automated. Even if 1% remains (such as agriculture), it’ll be good for us
• Should satisfy:
  1. Hard to automate
  2. Complimentary to AI (force multiplied by)
• Things not well-described in internet data / not RL-able
  • Physical tasks
  • High-level strategy
  • Long-horizon, open-ended, require teams
  • → knowledge work becomes top-heavy
• Where people demand a human in the loop (especially phyiscal)
  • Requiring legal liability
  • High reliability
  • Professional protections
  • Preference for human touch (eg. as a luxury)
  • Physical presence needed (eg. teachers, firefighters)
  • Institutional inertia (especially gov’t, monopolistic companies)
• Skills needed for AI deployment
  • Kinda like managing teams of humans
  • Complimentary skills like hardware, data-centers, cybersecurity
• Luxury goods (due to wealth inequality)
• Personal skills:
  1. Deploying AI to solve real problems
  2. Taking care of yourself
  3. Personal effectiveness + agency
  4. Prioritization, forecasting, decision making
  5. Social skills
  6. Learning how to learn
• Business skills:
  1. Leadership + management
  2. Operations management
  3. Policy + political skills
  4. Communications (including social media, events)
• Specialist skills:
  1. ML research + engineering
  2. AI governance + strategy
  3. Cyber + biosecurity
  4. Expertise in emerging power
     • China, multipolarity
  5. AI hardware
  • Also economics (especially growth + automation), public policy, applied math, robotics, engineering, philosophy, some history
  • Physical skills
  • Luxury services

9: Career Capital

• Big corporate tech + finance + consulting = “Bermuda triangle of talent”
• Better: smaller high-growth org

1. Take an established route into policy careers
   • Fellowships, working for a politician, think tanks, dep’t ladder
   • Meet people, say you want to find your way in
   • Write + publish about it
2. Graduate school
   • Do something relevant
3. Get a verifiable valuable skill
   • Eg. BlueDot
4. Do anything where you might excel, even if it’s random
5. Learn by doing

• And don’t do what others are doing

10: Finding the Right Career for You

• What could I become good at?
• Top performers are 100x better
• Generally just “pick what suits you best”
• Long-term career satisfaction = career capital + immediate impact + supportive conditions + personal fit
• Hard to predict (<40%)
• Shouldn’t trust gut - it can’t be trained for things like career choice
• Best: structured interviews + biographical data
• Process:
  • Rank all options in terms of fit
    • Especially hour-to-hour
  • List key uncertainties (that’d update rankings)
  • Devise cheap tests for ↑

11: When Should You Settle?

• Explore exploit tradeoff a
• Study: hot streaks tended to follow periods of exploration
• 4 reasons to explore:

1. Shoot for the stars
   • Upper confidence interval algorithm: start with highest upside option first
   • On average, people overestimate ability for small tasks but underestimate for big tasks like founding a company
2. Try out several paths, but with careful ordering
   • Traveling salesman problem
   • More costly than 1, so better early in career
   • Business school gap helps
3. Consider a wildcard
   • “Random” option
4. If in doubt, quit
   • Sunk cost fallacy
   • Status quo bias

• 5 factors for career path: (revised)
  1. Immediate impact
  2. Career capital
  3. Personal fit
  4. Supportive conditions
  5. Exploration value

4: Plan + Action

12: How to Make Career Decisions

• #1 factor for good decisions: make more options (narrow framing)
  • Eg. ask others / outside frame
• Job candidates best picked when according to objective criteria, so should do the same for careers
  • Narrow down to top 3-7 factors
• Initial rankings + key uncertainties
• Get data + action, jump once ranking settles
  • Pre-mortem + pre-party
• Make judgement call
  • Accept doors closed on futures
  • Listen to your gut as another input

13: How to Make a Career Plan

• “Plans are useless but planning is essential”
• 3 broad stages of career:
  1. Explore (18-24)
  2. Build career capital (24-34)
  3. Deploy (34+)
• Have a career vision: some broad idea where you’ll end up
  • 1-3 problems, 1-5 roles / titles
  • Best only for determining next steps
  • Work backwards and forwards
• ABZ plan: A + B + worst case scenario
• “Nothing in life is as important as you think it is while you are thinking about it” - Daniel Kahneman
• Avoid risks that may permanently reduce happiness or career capital, eg.
  • Burning out, getting depressed, ruining reputation, breaking important commitments
• Set a review plan / cadence

14: How to Actually Get a Job

• Job hunting is sales
  • Responsibility, salary, put their reputation on the line
• 3 key stages:
  1. Generating leads
  2. Converting employers
  3. Negotiating offers
• Could take 20-100+ leads to get a converting offer (6+ mo)
• Best jobs often at high-growth cos (bespoke + not on job boards) via referrals
• How-to*
• Do a portfolio project, case study, etc.
• Go for a nearby position → internal promotion
• Trial
• 3 key points, 1-2 key stories each
• 1-2 most impressive things you’ve done, not more

15: How to Network

• Surround yourself with the right people
• Networking: “meet people you like, help them out, and build genuine relationships”
• 5-min favor: intro, book, app rec
• Create content (in a niche)
• Host events
• Find your community
• Gains of trade often positive sum due to specialization / comparative advantage
• EA + AI communities best
  • Key: be part of 2+ to avoid lock-in

Final thought

• “On your deathbed, helping people is one of the things you’re least likely to regret”
• “If we really want to be satisfied, we have to stop focusing on ourselves and our own passions and instead turn our attention outwards to what we can do for other people”