A Country of Geniuses

“The interrogator would be unable to distinguish them.”
— Alan Turing, “Computing Machinery and Intelligence”

Dario Amodei asks us to imagine what happens when AI unlocks, "a country of geniuses in a datacenter." He asks us to imagine 50 million geniuses coming online.

Imagine we built it: A single machine the size of a nation. We called it The Middle Data Center, a black box of nearly boundless intelligence and labor, wired to a robotics superfactory that never slept and never asked for a raise. The bargain it offered was simple, and so it was irresistible: route your production into The Box, and The Box would return the finished goods, cheaper and faster than you could ever make them yourself. So we did.

We started with the easy things: the toys, the T-shirts, the cheap electronics nobody wanted to assemble by hand anyway. The Box did them flawlessly and for almost nothing, so we fed it more. Furniture. Appliances. Auto parts. Then the harder things, the things we'd been proud of: the steel, the circuit boards, the machine tools. Each time, a CFO somewhere ran the numbers and arrived at the only rational conclusion. Why keep a line running when The Machine did it for a third of the cost? The decision wasn't ideological. It was obvious.

And it worked. It worked spectacularly.

The quality-adjusted price of a television fell by roughly 98 percent. The price of toys fell by about 73 percent, even as the cost of nearly everything else in American life more than doubled. Apparel, furniture: flat for a generation while wages and rents climbed. Every family that shopped at a big-box store on a Saturday was quietly handed a raise they never saw on a pay stub. Hundreds of millions of people, a little better off, a little at a time.

The trouble is that abundance had another side to it: the people who lost their jobs when this machine of intelligence and labor came online. Savings were spread across everyone. But the losses were concentrated into specific places: small cities across the industrial Midwest. In those places The Box didn't deliver abundance. It delivered a termination letter.

The men and women who had made things found they couldn't retrain, relocate, reinvent. They slid into long-term unemployment and the slow gravity of public assistance. When the work went, the men who had done it became, in the cold language of the research, less "marriageable." Marriage rates fell. Fewer children were born, and more of the ones who were born grew up in harder homes. In the counties most exposed to the box, working-age men began dying in numbers that didn't fit any chart anyone wanted to publish. Overdoses, suicides, mostly, the deaths we eventually gave the term deaths of despair. The places that lost the most jobs lost the most lives. The two lines were basically one.

The abundance was real. The despair was also real. It just depended on who you spoke to.

Now here is the truth about the box: There was never any data center.

The Middle Data Center is the Middle Kingdom. China. On December 11, 2001, China joined the World Trade Organization, and the world plugged itself into a reservoir of intelligence and labor so vast it behaved, for all practical purposes, exactly like the machine in the thought experiment. Boundless capacity. Falling costs. A black box you fed your production into and from which finished goods returned, cheaper than you could believe.

Every statistic above is real, and it is about China, and it is about us. Economists named it the China Shock. They estimate the box eliminated somewhere on the order of two million American jobs in its first decade, and that the surge of imports accounted for more than half of the entire decline in American manufacturing employment over those years.

The China Shock has a lot to teach us about the AI Shock.

An Intelligence Explosion

I want you to sit inside the most uncomfortable implication of the whole exercise: We are not speculating about what happens when a society routes its productive economy into a black box of boundless intelligence and labor. We already ran that experiment.

We have the results.

We think that AI is different from everything that has come before us, because we have never been able to replace intelligence. But what we saw in China was an intelligence boom. It wasn't just the affordable robotic hands in this black box that have replaced our workers. It was also the massive intelligence capacity that has boomed onto the global stage, from the 100,000 engineers at BYD to the 996 culture (9:00 a.m. to 9:00 p.m., 6 days a week). Let's not kid ourselves: when we see swarms of Chinese robots doing back flips in perfect unison, it is the intelligence of its engineers doing that.

In 2003, the United States led in 60 of 64 critical technologies and China led in just three. By 2023, the two countries had switched places. China now ranks first in close to 90 percent of tracked technologies, spanning drones, satellites, advanced materials, nuclear energy and more. The concentration is staggering: in some fields all ten of the world's top research institutions are Chinese, collectively producing nine times more high-impact research than the second-place country.

China now graduates the entire engineering workforce of America every five years.

It was the analog version of an intelligence explosion, but it was an intelligence explosion.

The real reason that the semiconductor fabs aren't showing up faster in America isn't because we can't hire the labor. It's because we don't have people who know how to do it. That includes everything from specialized knowledge on how to run the machines, to the culture that lets us build fast enough, to the institutional intelligence to permit them. TSMC's American fab is a year behind schedule. They had to dispatch a roughly 500-person team of experienced technicians from Taiwan to Arizona to handle the tool installation and get back on schedule. You fly 500 people across the Pacific because the knowledge of how to actually do the thing lives in their hands, accumulated over decades in a dense cluster.

China was a beast of intelligence. Not silicon, but a billion people and a state that organized them. Dario talks about the 50 million geniuses in a data center. China statistically has about 30 million geniuses, all of whom are positioned to navigate the real world better than some agent in a box. A billion minds coming online, all ready to operate within the human system because they are the human system.

From the American checkout line, it was indistinguishable from a data center and a factory.

It is a great analog for what happens when abundant intelligence arrives, because it was a massive arrival of abundant intelligence. We should expect to see what has already happened in some parts of our economy happen to more of it.

So we don't have to imagine it. We can look at what it produced. And we can learn from it.

If we are going to be intelligent about placing our societal interventions to build a better future, we have to look for the clues where we can find them, and learn from them. This is as good a place to start as any. The entire debate is conducted in the future tense: what will happen if machines and intelligence are suddenly able to operate in abundance? What might become of the people whose labor they replace? But we are not standing at the edge of an unknown. We are standing at the edge of a rerun. For decades we lived alongside an entity that offered exactly the deal the AI optimists are offering now: hand us production and we will hand you abundance. The abundance came. And then a lot of other things did.

Round One: Lessons

What can we learn from this country full of geniuses and robots?

We have learned that abundance doesn't show up for those who can't afford it. We learned that Everyday Low Prices aren't a path to prosperity. That prosperity and poverty are concentrated in pockets. That our government cannot act in a coordinated, efficient manner to catch those who fall between the cracks. That swaths of a great nation can fall between those cracks, and cracks can become chasms.

We have learned that you can't just tax the trillionaires. Even if we taxed 100% of the paper value of Elon Musk's net worth, that would boost our budget by just 14% for just one year.

We have learned what it looks like when large numbers of people understand that they are irrelevant. We have learned that those who don't participate in this prosperity begin to decay, not just economically but in what they eat, in their health, in their energy levels, academically, in how hard they work because of how much hope they have. They decay in their very culture. We have learned that these lives of despair lead to deaths of despair.

We've learned that it's hard to help people in families and communities that have fallen into this state of disrepair, because family and environment matters more than any intervention. We have learned that we as American society have not learned how to make it work.

Who wins in this world? Yes, the consumers. Their gains were real. For the ones with jobs.

But we also learned that when some things get cheap, other things get more expensive. Progress changes the shape of scarcity. The goods that can be copied, shipped, scaled, or manufactured collapse in price. But prices in other parts where humans are required become more expensive. Economists call this Baumol’s Cost Disease: a string quartet still requires four musicians playing for the same amount of time, no matter how productive the rest of the economy becomes, and so they become more expensive. It’s the reason we don’t have artisans building facades. Costs tied to human presence, trust, care, land, institutions, and time can rise wildly. So the costs of education and health care skyrocketed while the cost of TVs and toys plummeted. We should expect the Box to do this again. It will make some things abundant beyond imagination, while making other bottlenecks more valuable, more contested, more scarce. Those who are prosperous will buy their way through the new scarcities. Those who are not will be told to be grateful for the cheaper things.

But the people who gain, who really gain, power and money or both, are the ones who control the box: the capital that decides where production should go, the technology that it runs on, or the factories that do the producing. For most of the last century, before The Box, the share of national income that went to workers as wages had sat stubbornly around two-thirds, so steady that economists treated it as a law of nature. In the decades The Box operated, that share fell. The income it shed did not spread back across the workforce. It pooled. A small number of dominant, capital-rich, technology and financial firms captured a swelling share of every market while paying out less and less of their value to the people who worked for them. This is the world of trillionaires.

Monopoly was invented to warn us. Its creator, a radical opposed to the way capitalism let wealth pile up in a few hands, designed the game to show that under those rules money always aggregates to whoever already has the most power. She meant it as a cautionary tale. We turned it into family game night in a capitalist society. And in this system, where the power has gone to the people who sit atop The Box, it turns out she was right.

We have also learned that the countries that benefit are the ones that build enduring skills, network effects, companies, and cultures that thrive, because they run and own The Box. We've learned that the people who somehow become part of this engine and its success, either because they own it, contribute to the technical capability that pushes its frontier, work within it, or are the capital behind it, are the people who thrive. They are so successful, so rich, and so powerful that they become the envy and the object of anger of all of the people who are not.

Who is positioned to own The Box the second time around? The logic that pooled the gains last time is the logic AI runs on by nature: enormous fixed costs to build, almost nothing to copy, a winner-take-most race that rewards whoever already has the most. The capital is the hundreds of billions being poured into compute. The technology is the handful of frontier models. The factories are the data centers. Whoever controls those three things stands exactly where the owners of the box stood twenty-five years ago, except this time The Box does not sit across an ocean stitching T-shirts. It sits in your office and your bedroom.

So when someone asks me what I think AI will do to the future of work, I think about the cheaper television and the emptier Main Street and the savings spread thin across everyone and the ruin poured thick into a few. I think about how real the gift was, and how real the grief was, and how we never managed to get the gift to the people who paid for it. I sure wish that we as society had looked ahead then and asked ourselves: How do we train the young in these places to thrive, while we still had the money to do it, the united families that valued hard work, and while they were young enough that they were easy to train.

The box always delivers. It delivered cheaper goods and emptier towns; it handed a refund to everyone and a fortune to a few. The two questions we never answered the first time are the two we are about to be asked again: not whether the box works, because it always works, but who ends up consumed by it, and who gets to create it, benefit from it and own it.

This Time Is Different

This time is different, they say. Yes. It always is.

This time is different, we have gunships and gunpowder to conquer the world. This time is different, we have railways and machines and steel to build a nation. This time is different, we have electricity. This time is different, we have computers. This time is different, we have artificial intelligence and robots. The technology always changes and, yes, this one may be the biggest of them all. But the laws of civilization and human nature never do.

I don't buy the argument that the Black Box of AI will produce so much abundance in any of our lifetimes that people won't know what to do in the luxury of their homes with the luxury of all of their free time, any more than I now buy the argument that offshoring America's core competency in manufacturing was going to lead to universal abundance. Yes, it produced productivity and incredible gains for many. But unemployed people cannot afford the basics, even if the basics cost less. We should expect that what happened with the last black box will happen again, faster and bigger, since a bigger and better Box is coming.

We have learned from all of history that durable prosperity is broad prosperity. That the societies which stay rich are the ones where large numbers of people can both contribute to and benefit from the Engine of Prosperity, and that when the gains pool at the top, decline is already underway.

We assume the future will be so different we can't even guess at its shape. We assume it is destiny, that we either have no choice but to give up, or that it will all just work out fine.

But we already ran the experiment. We already know what it's like to bring online a nation full of geniuses and superfactories. The only path that stands a chance of working is to ourselves become a nation full of geniuses, all empowered by that nation full of geniuses.

What does it look like to do that? Let's turn, again, to China.