I'm a libertarian, but the AI era might need a big government

A few nights ago I couldn't sleep, and while skimming through AI news in the early hours, I got a strange feeling.

I've always been on the side that values freedom. I don't like the state intervening too much in individual choices. I've believed the better way is for individuals and companies to compete freely in the market and find answers that way. I thought government was better off focusing on the places it's genuinely needed rather than trying to do too much itself.

But watching how AI keeps rolling forward, my thinking has gotten uncomfortable lately.

Is small government really the best government in the AI era too?

I wonder if maybe the more technology advances, the more there is for government to do.

An even stranger thought comes to mind. What if solving the problems created by capitalism and technology working too well ends up requiring far stronger redistribution than we have now?

AI isn't a story that ends with one new technology

The world changed a lot when the internet arrived. Same with smartphones. But the change I feel watching AI has a different grain to it.

That's because AI is less a technology that transforms one industry and more a technology that lowers the cost of human intellectual labor across the board.

It does part of what developers did. It does what designers did. It translates, researches, analyzes, writes documents. Bolt it onto robots and it crosses over into physical labor too.

I tried a rough thought experiment.

Say there's a company that used to need 1,000 people. With AI and automation attached, 100 people produce the same output. From the company's perspective, this is remarkable. Productivity goes up, costs go down, profits go up. Viewed across society as a whole, it's technological progress.

One problem remains.

What happens to the other 900?

Rising productivity doesn't mean everyone's income rises

Since the Industrial Revolution, we've accumulated the experience that technological progress makes society as a whole richer. Over the long run, that has actually been true.

But how the fruits of the productivity gains AI creates get divided, and along which paths, is a different question.

The companies holding the AI models. The companies holding the data. The companies holding the computing infrastructure. The companies making the semiconductors. The companies making the robots. And the people holding equity in those companies.

The smaller human labor's share of the means of production gets, the more wealth attaches to capital rather than labor.

That creates a strange situation. Society's productive capacity is the highest it's ever been in history, while the labor value of a great many individuals actually falls.

The world got richer, but the money to buy that wealth isn't in most people's hands.

Technically it succeeded, but the social system threw a fault. That's the picture that bothers me most.

The paradox of capitalism working so well that redistribution becomes necessary

This is the part I find most ironic.

AI isn't a technology built by a state-run planned economy. Fierce competition among companies and enormous private capital produced this speed. It's classic capitalist force.

But if that competition succeeds and AI's productivity climbs to the extreme, society ends up demanding things like this instead: a thicker social safety net, job transition support, education support, social insurance. In some cases, even new income distribution schemes like basic income.

Draw the picture and it looks like this.

AI/automation → productivity increase → capital income increase → taxes → redistribution → consumption sustained

It's odd. As a result of capitalism working out too well technically, keeping capitalism running requires stronger redistribution.

That said, calling this socialism outright isn't accurate. State ownership of the means of production and maintaining a market economy while thickening taxes and the social safety net are different things.

But it clearly creates a point of friction with the traditional small-government intuition.

Looking at China brings up another uncomfortable thought

This is also where my sense that China is formidable in the AI race comes from.

AI isn't an industry you compete in with one good algorithm. You need semiconductors, you need enormous amounts of electricity, you need data centers, you need networks. Robotics and manufacturing get attached, education policy connects in, data policy gets caught up in it, and it runs all the way to national security.

AI competition increasingly demands a nation's whole-system integration capability.

On this kind of board, the Chinese-style system — where the state can designate strategic industries and push capital and infrastructure in one direction — works fairly powerfully in certain areas. The speed of laying power grids, building factories, standing up data centers, attaching industrial policy, and getting companies moving in one direction. Here, the advantages of a centralized system are clear.

For a liberal like me, that's a fairly uncomfortable advantage.

That doesn't mean the Chinese model is the answer

Moving fast and moving in the right direction are different problems.

Centralized systems decide fast and concentrate resources all at once. In exchange, the price they pay when a decision is wrong is also large.

Frame it as a system and it clicks quickly. Say one powerful Control Plane moves a large number of Workers at once. When the Control Plane's judgment is accurate, the speed is astonishing.

And when the judgment is wrong?

The Blast Radius becomes the entire system too. It doesn't take long for one bad configuration to spread to every node. I've seen this scene a few times in operations, and each time I relearned what the price of speed is.

The free market economy sometimes looks inefficient to the point of being frustrating. Countless companies develop similar things, investment gets duplicated, some companies fail. Government and business have different interests, and decisions are slow.

But inside that chaos, independent experiments run simultaneously along multiple branches. When one side is wrong, another tries a different approach. Something nobody at the center thought of pops up from below. That's the strength of a distributed system.

Small government or big government — that question might be outdated

So lately I wonder whether the question itself needs to change.

The old debate went like this. Big government, or small government?

In the AI era, a different question comes to the front. Not how big the government is, but how well it's designed.

Regulate AI too heavily and innovation dies. Do nothing and technology and capital pile up in a handful of companies. Overdo redistribution and the reason to build something new disappears; make it too weak and the enormous productivity gains AI created come back as social conflict. If the state controls data too tightly, individual freedom gets shaved away; leave it completely alone and giant platform companies end up holding de facto power over data.

Neither side is simple.

Maybe government is an architecture problem too

After developing long enough, you don't start counting servers when you look at a system. You look at how far to centralize and how far to distribute, who holds which permissions, how far a failure spreads, what gets automated and what a human has to approve.

I wonder if you can look at a country in a similar way.

A system where everything is decided centrally is dangerous, and a system where everything is left to the market also collapses under certain conditions. What matters is putting responsibility and authority in the right places.

Seen that way, national competitiveness in the AI era shifts from being a question of government size to a question of how well you design the nation's architecture.

We might need better government precisely to protect freedom

I still think freedom matters. I think a society that gives individuals choices, lets companies take on challenges freely, and even permits the freedom to fail is a good one.

But paradoxically, in the AI era, protecting that freedom may require government to become more sophisticated.

Blocking monopolies, watching whether the wealth AI creates is piling up on one side, giving people displaced by technology a chance to compete again, and at the same time protecting the freedom of the people building the new technology.

What I want isn't big government. But it isn't unconditionally small government either.

A government that's strong but doesn't put its hands everywhere. One that steps back where it should step back, and intervenes precisely where the system would otherwise break. Easy to say, but it's probably the hardest kind of government to build.

The more AI advances, the more the gap in this capability may widen alongside the technology race itself.

Competition in the AI era won't just be a fight between American AI and Chinese AI. It won't just be company versus company either. It'll be about which society best coordinates technological innovation, individual freedom, and the distribution of the wealth that technology creates.

Everyone's watching the performance race between AI models right now, but behind it, a much bigger competition may already have started.

What is the operating system for a nation that best fits the AI era?

I don't know the answer yet.

But one thing has gotten harder to be confident about than before. The idea that as technology advances, government will naturally get smaller. It might be exactly the opposite.

The more powerful technology gets, what we need may not be more government, but far better-designed government.

Even after writing all this, I'm still half a liberal. I haven't found a name for the other half yet.