stop trying to be a "good person"

the bitter lesson of consequentialism.

[epistemic status - I am quite confident all of this is true¹ and every rebuttal I can think of is handled in a footnote]

UPDATED epistemic status: Excellent critcism in the comments, lots of use of hyperbole and non-arguments in the article. I also presuppose utilitarianism. epistemic status is neurondumpy frustration at best.²

This post was sparked after I met someone who earns a lot of money, and donates around ~$200K+ every year to effective animal charities. One of my friends got really annoyed with me after I called her “probably the most ethical person I know”. But today I want to double down on that. What is being a “good person” if not doing the most good?

---

Bad intuition

If you ask most people what “being a good person means” they will respond something like “be kind, love thy neighbour, don’t steal, etc etc…”

If you care about impact, our intuitions regarding goodness just kind of suck: When you ask most people to picture someone environmentally conscious they will conjure up an image of a garden, a detached home, maybe an electric car, buying local, etc…

The real most environmentally conscious person³ is probably someone working for Hudson River Trading who donates 500k a year to climate offsets.

Like, yes: going vegan, living in a small apartment in a big city, not driving cars, living frugally do reduce your personal climate impact, but you could do just as much good on the carnivore diet, driving a gas guzzler and starting a campfire every day by donating as little as ~$40 monthly.⁴

Unless you’re a genius or a saint, your money is the strongest tool you have to change the world.

I think this is part of my frustration I had at EAGx Amsterdam. I saw so many people who wanted a “job in EA”. They wanted to directly do the good. Have they really thought through the bitter truth? Why do you believe you are uniquely good at an EA job, why ignore the simple premise of earning to give?

The bitter lesson

You get really good at something, get a high paying job in the field, and then donate to causes you think matter.⁵ For most people, nothing else matters much.⁶

It should hardly be a surprise that earning to give works so well, we live under an economic system that prides itself on how you can make 1000 times more bobby pins, as long as everyone specializes. And then we ignore this premise when it comes to doing good, and support perfectly high earning finance people pivoting their careers to AI safety!⁷

If you can do good with your career AND give a lot, that is awesome, but I think for most of us, simply earning to give is seriously underrated. It’s boring, and not sexy, but it’s probably true.

Let’s say you live in London and work in tech. Your yearly expenditures are around $45k. Your wage, after tax is probably around $55k, but if you put in the hours and “get good”, it could probably get it to ~$100k⁸. In the long tail, it could probably be closer to $200k.

For each of these income levels your donateable income is⁹ $10k, $55k and $155k. Notice how much getting good pays off here! Going from $55k income to $100k means donating 5 times as much! Going from $100k to $200k means donating 3 times as much again!¹⁰

Simply pick something, get good, get paid, donate, gatorade.¹¹ Luckily, this is also a great recipe for a happy life.

I also think this extends to personal attitudes by the way. Going personally strictly vegan¹² is probably not very impactful at all. Farmkind has estimated it at $22 a month (which I take issue with!). That’s around $300 a year. The average annual raise is 10 times higher than that, so it’s probably easiest to just pursue that instead.

You have 80000 hours to spend on your career, dont spend them agonizing about what to eat, how many flights you take and how “good of a person” you are. Keep your eyes on the ball.

There is something to be said for not turning into a complete dickhead, and for personally going vegan. Because it probably prevents value drift and you know, helps social cohesion and all that, and getting out of bed every morning to save the chickens is probably easier if you are not eating them.

The incentives are pretty well aligned regarding being kind and doing good for money related reasons. You will not get far career wise by being an asshole to people.

Money makes the world go round. There’s this entire group that just studies how you can do the most good you can with it. Have you considered simply trusting them?

Second order consequences

“But the second order consequences of working in finance and perpetuating evil capitalism!” I hear you cry. Or maybe you thought of a better second order consequence.

Second order consequences are real. It’s why I don’t think doctors should go around stealing organs. The problem arises when this hijacks motivated reasoning and you hide behind the idea of “second order consequences” without thinking.

For any second order you would bring up, I would basically have the same answer.

How confident are you that there is a counterfactual that has more impact than a life spent saving 400 children from blindness and malaria, providing a wage for someone people to work on existential risk while saving millions upon millions of animals from living gruesome lives and dying gruesome deaths?¹³ What does this counterfactual look like? How confident are you that these changes would be better? And how much does working in finance prevent someone from actively supporting these causes? And why can they not do it with money? What is its track record? And how sure are you? Because the EAs are pretty sure. (if it passes all these tests please comment down below, I want to hear it!)

So yeah, as EA grows, I think it’s important that we reiterate to new members that donating is a simple and very effective way to do good. Not everyone should have a sexy EA job!

If you found this at all convincing: please consider signing the giving what we can pledge. If you’re as lucky as me you can even get a free lifetime ACX subscription to go with it.

1. though I basically presuppose utilitarianism

2. I think there are arguments for earning to give being overindexed on, but I do not think this piece argues very well for them. That aside, this might be fun reading for EAs. Leaving this up because it’s my most popular piece

3. (aside from being born into wealth)

4. (climate wise). also claude math didn’t bother to check. Exact number doesn’t even matter that much, points stands. It’s low

5. The name of this blogpost is derived from the famous Richard Sutton essay. If you don’t know who this is, he invented reinforcement learning. The essay goes as follows

1. humans try to figure out human-centric ways to make models learn faster.

2. Compute scales, years later someone comes along and goes “have we tried just making an architecture that generalizes well, throwing out all the other tricks and throwing a shit ton of compute at it”

3. this guy wins awards.

I don’t think it neatly maps or anything, but it rhymes

6. Cases where this is not true:

1. You are in fact a genius, or so good at something that you can single handedly change the world. But since you are reading celeste-land I’m gonna assume this isn’t the case, sorry

Now this is probably not true for everyone, sometimes your skills are hard to monetize or you are just exceedingly good at making shit happen. But that is not the majority of us.

7. A bit ragebaity, AI safety is kind of an exception to all of this, since if you believe it matters at all you probably believes it matters a lot, and it’s in need of very good people in a way that money can’t simply solve, quite urgently too.

8. Median wage in tech vs average levels.fyi wage for berlin. Assuming levels.fyi selects for cracked people and not taking a higher percentile because that’s selecting for seniors and insane outliers

9. ignoring tax exemptions

10. admittedly london has high salary caps and is competitive so the scenario isn’t entirely realistic. But hey, if I really wanted to cherry pick I could’ve picked SF x)

11. get laid (optional)

12. except when it comes to complicated second order effects of turning others vegan, which I’ve written about here

13. projecting out lifelong donating $80k/yr for 50 years, quite a high estimate but not completely unreasonable.