The turbulent AI era is here. The choices we make now are critical.

I'm going to make an anti-capitalist argument in order to make a pro-capitalist argument.

The anti-capitalist argument is that the wealthiest people in the world are not dramatically more intelligent than the average person. Bill Gates is obviously very intelligent and capable, but his IQ does not scale with his wealth. He built the right company at the right time, and Microsoft ended up sitting at the center of massive network effects. Relatively modest advantages in judgment, skill, and industriousness can end up producing enormous differences in wealth due to the compounding effects of success.

That's just how scalable markets work. So Gates's success, while evidence that he was very effective at building Microsoft, is not evidence that he is vastly more knowledgeable than everybody else about economics or the future of technology.

And the same applies to Bezos, Musk, and every other extremely successful founder.

Their arguments should be judged on their merits and not treated as infallible insights from people with superhuman intelligence.

Now, that is not to say that I want their wealth to be capped or politically reallocated. The case for private property depends on the argument that decentralized ownership and investment and capital risking is better than politicians allocating a pool of resources to gain clout or win over popular opinion. A wealth cap or high taxes on the wealthy would transfer control over resources from private owners to political institutions. So what we would be doing is replacing imperfect private allocation with imperfect public allocation, where the people allocating the capital have limited information, political incentives, and the ability to impose costs on others without their consent. So I reject blind deference to billionaires, but I do not want politicians to control billionaires' capital.

Anyway, let me address Gates's argument about AI:

Gates argues that AI will disrupt the economy faster than previous technologies, and therefore stronger government coordination is needed to address the rapidity of the speed of that transition, to cushion people against the speed of that transition. But what his analysis underweights is how quickly AI can help people adapt to the very transformation that AI brings about.

For example, a worker taking on a new position now has access to a tutor and an analyst and a business advisor all wrapped up in one, who can provide world-class assistance 24/7 at the cost of a few cents in electricity. And someone who wants to start a business can employ dozens of AI agents at a cost of a few hundred dollars to test out the idea and build out a prototype within a few days. So what this means is that we cannot hold the human adaptive capacity constant while trying to determine the impact of AI's disruptive capacity.

The second problem with Gates's argument is that he only counts the income that people might lose when AI replaces jobs, but he doesn't count the fall in prices that the same AI will bring about. The fall in prices is effectively an increase in purchasing power. That increase in purchasing power, in economic terms, means an increase in real wages, and that's exactly why wages increased twentyfold between 1826 and 2026. Automation brought about abundance, which reduced prices and thereby increased purchasing power, i.e. increased real wages. If you look at tokens right now, their prices are going down exponentially. They've gone down about a thousandfold over the last four years, while the capability of the tokens themselves has increased. So AI is following exactly the same price trends as every other previous computer technology.

If AI becomes a major input into medicine and education and administration and logistics, that means that all these goods and services become radically cheaper, which means that paychecks can buy a lot more of them. So that's, I think, a huge hole in his analysis. It's also, I think, very simplistic to assume that the future will be based on labor. It could very well be one where the average person becomes a business owner and employs a fleet of hundreds or thousands of AI agents to work for them. That's exactly what the price trends in tokens would suggest.

The broader problem I see with his model is that it assumes disaster from a technology that creates abundance. This seems to me to be fundamentally misguided about the essential dynamics of what's actually happening.

Finally, if AI is ever capable of doing every single thing that a human can do, AI would possess very dangerous traits. And why I say that is because humans themselves are very dangerous. If we're going to create a product, we're not going to create a product that has human-like traits. But it is some of those very traits that make humans dangerous that also make them useful in the economy. We tolerate humans having these dangerous properties because that's what we were born into, and we accept that and we legally protect humans despite the fact that they're dangerous. But nobody, no company, no country will tolerate people creating AI that is dangerous. An autonomous, self-motivated and potentially rebellious digital intelligence is not only not a useful business tool, it's also not a tolerable risk. And if that is ever created, we will have much bigger things to worry about than unemployment and no solution to the problems that this would create would be found in greater taxation.

And as long as that is not created, then there will be abilities or skills that humans uniquely possess that will become the limiting factor in the economy. In other words, all other processes will scale exponentially through mass production of GPUs until all the available scarce human skills are employed in managing those GPUs. That follows the basic pattern that has preceded us over the last 200 years, and there's no reason to assume that pattern will not hold.