In college, I temped for a headhunting firm. They taught me to spam engineers with LinkedIn messages, to recruit for an app called Spoon. I sucked at this, because I hate bothering people. (I didn’t know what headhunting was when I took the job.)

I customized every message for a human touch, ignored anyone who looked busy in their profile picture, and reached a third as many people as the other temp. Together, we convinced about six engineers to let us call them.

Spoon’s CEO was not happy. To make a point, he spent an hour sending his own messages, which were brief and clearly written by an engineer. He got more bites than we did in a week. 

Unfortunately, he didn’t have much time to recruit; he was too busy being a CEO. Someone good enough to impress an engineer through LinkedIn can use their time for better things.

This is what Patrick McKenzie calls “The Sort”:

The Sort is the world’s way of moving people to wherever their talents will produce the most money. It sends mathematicians to Jane Street, Dominicans to the MLB, and Justin Bieber to Universal Music Group.

But it doesn’t just show up in the economy. It’s also there in the curse of cryonics:

If something is important but doesn’t pay well, it’s hard to get people with expensive skills (or even the cheaper skill of “noticing important problems”) to do it. Cryonics could conceivably compete for talent with the tech industry, because it pays in expected utility and the glow of a holy mission — but AI safety crushes it on both fronts.¹

My employer, 80,000 Hours, is at war with the Sort. We take people who could earn millions of dollars in a frontier lab or corporate law firm and nudge them toward jobs that pay much less but might save the world.

We also have to fight the internal Sort that misallocates talent even within our AI safety bubble. I’m writing a profile on fieldbuilding — the unsexy work of developing talent and building infrastructure to help the field grow. It’s a lot like tech recruiting: the best candidates are AI researchers and ace networkers, but the researchers want to write papers and the networkers want to write policy. Our current strategy is to yell “please do fieldbuilding — nine out of ten leaders say it’s more impactful!” and hope that people listen.

Ways to get around the Sort:

Pay more money. This is the obvious and most visible answer. You see it whenever the smartest college grads flock to a new field. Trad finance -> quant finance -> AI. 

Unfortunately, doing this requires enough money to outbid the companies that already bought the smartest people and put them to work making more money.

Add a sprinkling of prestige. When 80,000 Hours and Coefficient Giving yell about fieldbuilding, we’re making a rational argument about impact — and also sending a signal that trustworthy, competent organizations will be proud if you run conferences instead of writing papers (in a world where papers are considered “cool”). 

On a grander scale, Teach for America got quite a few Yale grads to take ordinary teaching jobs — though few stuck around.²

Make the work more fun. Many professional Magic: the Gathering players also play poker. I’d guess that poker pays between ten and infinity times as much per hour. But Magic is more fun, so the pros trade lucrative time in fancy casinos for sub-minimum-wage time in conference centers with concrete floors.

The best fanfiction is considerably better than Amazon’s top fantasy novels, especially if you correct for the absence of editors. Nonetheless, its authors write for free: HarperCollins won’t let you put Naruto in your story (or show him in a three-way with Hinata and Sakura).

Dig deeper for talent. Every so often, society cracks open a new source of human talent. It sends everyone to school; starts hiring women; builds fútbol academies in Brazil and chip factories in Taiwan. 

For a while, it looked like our next expansion pack would come from ending lead poisoning and giving kids the right amount of iodine. But artificial brains came faster. We’re about to have all the talent we can use; I hope it doesn’t Sort us out of existence.

There are some people the Sort can’t reach.

My wife is a cancer surgeon. Most cancer surgeons do the job because they love removing cancer. As long as the patient survives, they’re satisfied. 

My wife prefers the art of reconstruction: putting patients back together with as much function as possible. (This week, she returned an ear to its rightful place and gave someone a new nostril.) Most people like her wind up doing facial plastics, which pays better than surgical oncology and gets you home earlier. But she likes saving lives too much to give up on cancer.

I don’t know what the world can learn from my wife. She’s unrelentingly stubborn and insanely dedicated to healing — the person who helped me found an effective altruism club and then almost boycotted the whole concept because 80,000 Hours told her not to be a doctor. But even if she doesn’t generalize, it’s pretty cool to watch her shrug off one of the century’s primary sociological phenomena through sheer willpower. Love you, Tammy!

1. Recently, someone wrote LessWrong’s top cryonics post after a seeming scientific breakthrough. I don’t see this as strong evidence against the curse, because (a) the rate of cryonics posts still dropped after GPT-2 and (b) the author’s company started 10 years ago. I’d be somewhat surprised to see a LessWrong user launch a new cryonics firm today.
2. I only knew one of them personally; he works on AI safety now.