Reflections on AI

I first heard of this essay during Ezra Klein’s interview with Sam Altman from 2021. From Klein’s description of it, I honestly expected the essay to be mostly about the ways in which Altman imagines that AI will transform society for the better. I understood that the general premise was that AI would unlock a phenomenon like Moore’s Law for nearly every sector of the economy; i.e. all forms of labor would become exponentially cheaper every year thanks to assistance from AI, to the point that anything you can imagine that was previously expensive to build or achieve would be democratized.

What I didn’t expect is that the essay would be so light on details about how AI would transform society, while spending much more time proposing tax policy ideas.

The essay is divided neatly into 5 parts. I’ll summarize them here:

1. The AI Revolution. Altman asserts that we are entering the fourth major technological revolution, following the agricultural, industrial, and computational revolutions (I think this last one is often referred to as the “information age”). He says more progress will be made in the next 100 years than everything humanity has achieved up until now.
2. Moore’s Law for Everything. Altman predicts that thanks to AI, we will reach a point when everything is getting 50% cheaper every 2 years.
3. Capitalism for Everyone. Altman predicts that taxing income (which he equates with taxing labor) will become increasingly inappropriate and ineffective as the AI revolution transforms the economy. He proposes a new nation-wide equity fund, based on taxing capital (of companies over a certain size) and privately-held land, and distributing it to all citizens over 18. He argues that this will align citizens’ incentives with wanting the country as a whole to thrive which will in turn benefit everyone.
4. Implementation and Troubleshooting. In this section Altman gets specific by doing some napkin math and estimating that a decade after implementation, all American citizens would be receiving $13,500 annually from the nation-wide equity fund, and that this would provide significantly more purchasing power than it does today (thanks to the price of everything having decreased significantly during that time). He acknowledges some challenges that would likely arise along the way and suggests ways to address them.
5. Shifting to the New System. Finally, Altman notes that making the changes he is proposing all at once would cause some amount of shock; so he suggests a more gradual approach that ties the new taxes on capital and land to GDP growth (tax rates are increased incrementally each year until GDP has increased 50% from when they were first implemented). He believes that policymakers will be very popular for supporting this system. Altman concludes by saying, “The future can be almost unimaginably great.”

I have mixed feelings about this essay. On the one hand, who can argue with the future being unimaginably great? I also really appreciate that Altman took the time to “do the math” on his proposal.

On the other hand, I really do wish he spent more time exploring the ways that AI will concretely make the world better. He gives the example of robots building a house, and points out that even the robots would be extremely cheap if they were made by other robots. That’s probably the most specific example he gives. In my opinion, more specific ideas would go a long way towards getting the reader excited about the future he believes AI will bring, which I think is at least part of the point of the essay.

Without effectively conveying why we should all be so excited about AI (my gut says that the majority of readers either already feel this way or don’t; the essay won’t change a lot of minds), Altman gets a little too hand-wavy with his “Capitalism for Everyone” proposal, in my opinion. The rough math seems to kind of work out, but there are so many unanswered questions. One example for me would be how we should think about the purchasing power of $13,500 after a decade compared to the median income of someone today who will be effectively unemployed in the future. Is Altman confident that Moore’s Law for Everything will make everything so inexpensive that it more than offsets people’s loss of income due to drastic decreases in the cost of labor? The force driving prices down would be the same force driving wages down and unemployment up.

In fairness, Altman does describe his entire proposal as a conversation starter; so I don’t think he believes he has all the answers. Overall I do think there’s a lot to like in what he’s saying. But given the risks if things go wrong, he could probably spend a bit more energy on making a really airtight case that AI will do (far) more good than harm.