Don't Listen to Sam or Dario

Sam and Dario have no idea what AI will do to the economy or society.

They're predicting mass unemployment. Should we listen to them? Absolutely not.

Stay in your lane, boys.

1. The inventors of new technologies have always been hilariously bad at predicting their consequences.
2. They're not alone: almost everyone is bad at predicting the future, except one weird group.
3. The early evidence says Sam and Dario's predictions are dead wrong.

The inventors are always wrong

Most inventors have no idea what they're talking about. The inventor of the Segway thought it would replace the car; all it's done is become the official ride of mall cops. Even when they're right that their invention will change the world, they rarely predict how.

• Edison thought his phonograph would be for dictation, not music. And when talking movies showed up, he insisted Americans would never go for them.
• Richard Gatling built the machine gun to shrink war. It just made it more horrific.
• Zuck thought Facebook would lead to more human connection. It precipitated a crisis of loneliness and the occasional genocide.

There's a simple explanation: technologists are weird, and society is a dynamic system. Extrapolate from your own unusual perspective, fail to account for how the world absorbs and adapts to a new technology, and your prediction becomes a punchline.

Listen to inventors on the capabilities of their technology. Never on its consequences.

Almost everyone else is wrong too

Philip Tetlock has spent forty years asking people to predict the future, and then scoring them.

His biggest finding: almost everyone is wrong, almost all the time. And the more famous the forecaster, the worse the forecasts.

Sorry, Sam and Dario. The fact that I don't need last names to identify you is a priori evidence that you're too famous to listen to.

But Tetlock found a small group of people who are provably good at predicting the future. They're called superforecasters. In peer-reviewed studies, they were better than CIA analysts with access to top-secret data.

So what do these oracles think will happen with AI? Do they agree with Sam and Dario? Their verdict:

• Full employment
• Modest GDP growth
• The models keep getting smarter

In other words: the only group of people who are actually good at predicting the future think AI will keep getting better and make the world mildly better along the way. Their P(doom) is basically zero.

The early evidence says Sam and Dario are dead wrong

Let's look at some specific predictions from AI leaders a few years ago and see how they did.

The narrative: radiologists are toast. The reality: radiologists make $570,000 a year, and there's a shortage.

In 2016, Geoffrey Hinton said, "People should stop training radiologists now." Ten years later, radiology pay is rising faster than almost every other specialty's. AI reads the scans better, but radiologists do more than read scans, and when the scans get better, people order more of them.

The narrative: software engineers are in big trouble. The reality: job postings are up.

Dario said AI would soon write 90 percent of the code. He was right. It does. He was wrong about what it would mean: the more AI-exposed the job, the faster its postings are growing. People wanted a lot more code than they could buy before. My dad's janitorial company has an engineer now. Of course it does.

The narrative: AI will destroy lawyers. The reality: law firm hiring is up, and hourly rates are up.

The class of 2024 set an all-time record for law-grad employment, and billing rates jumped almost 10 percent last year. That first-year associate who was useless their first year is pretty damn useful paired with Claude.

The narrative: entry-level white-collar work dies first. The reality: the few companies that actually fired the kids are hiring them back.

In 2024, Klarna bragged that its AI was doing the work of 700 customer-service agents. Fifteen months later, the CEO was rehiring humans, because the AI-only service was "lower quality." Meanwhile, youth unemployment sits around 9 percent, below its long-term average. New-grad unemployment is the one elevated number, and the New York Fed blames remote work, not AI.

One real casualty so far: US customer-service reps, down only 10 percent since 2022. They're this revolution's buggy-whip makers. And even that will be a blip: the US destroys tens of millions of jobs and closes one in ten businesses every single year. Churn is the baseline.

Here's my bet: more people will get raises than pink slips, and more people will quit to start their own businesses than get laid off to automation.

This is a golden age. People are just determined to blackpill.