Elon Musk leads Grok and Optimus, a frontier large language model and a humanoid robot. From his vantage point, he reports that "in a benign scenario, probably none of us will have a job." Yet he's not concerned: "There will be universal high income."
The expectation that mass unemployment awaits, and that everyone will get a handsome paycheck whether they're working or not, has been floated by AI's other leading lights: Dario Amodei (Claude), Demis Hassabis (Gemini), and Sam Altman (ChatGPT).
The case for mass unemployment almost makes itself. Once an AI can do jobs 100 times faster for one hundredth of the cost, surging unemployment won't be much of a puzzle.
But the promise of a "universal high income" sounds absurd on its face: Unemployment benefits today consume about 1% of the national budget, whereas giving everyone even a miserly check would consume over 100% of the national budget. If a subsistence wage is such a non-starter, the promise of a universal high income looks like outright charlatanry.
So what, exactly, are Musk and his frienemies thinking?
Just this: what's unworkable today may be doable when AI replaces us.
Imagine you have an identical-twin robot that you send to the office to work in your stead. Your boss doesn't notice, so your paycheck keeps coming, but she notes your performance improving. Meanwhile, you're at home with time for family, exercise, and sleep. Everyone is better off.
You're not working, still drawing a wage, and no one is out of pocket!? Yes! Paying you if you've been displaced isn't affordable, but paying you if you've been replaced by a faster, smarter digital doppelgänger, no less, is a bargain all around. And what is true for you would be true for the economy writ large: when we're all replaced by more productive technology, there's enough money in the system to pay everyone well. That's what Musk & Co. have intuited.
So the "benign scenario" isn't smoke and mirrors; the math works, but it won't materialize spontaneously. Left to market forces, expect employers to pocket your paycheck and organized labour to fight back. The likely outcome? Capital wins, labour pays the price, and the conflict hampers AI adoption during the critical phase of the global race, leaving everyone worse off.
Yet if policymakers act now, a better outcome is available. What's needed is a mechanism that distributes the new abundance automatically, without waiting for a crisis, and without relying on goodwill. That's what the MOSAIC Model provides.
The first pillar is a "negative income tax." NIT is a rare point of consensus among economists on both the right and left (Milton Friedman, Friedrich Hayek, James Tobin): just as today we pay income tax only if we earn above a certain threshold, in the future, income tax would pay us if we earn below a certain threshold. This creates an 'income floor': anyone without a salary receives a top-up to that floor; as earnings rise, the top-up tapers, but work always pays.
Some income floor is theoretically possible today, except there isn't enough money to set it at a meaningful level. Which brings us to the second pillar: mechanisms that raise the income floor as AI abundance actually materializes.
Start by following the money.
When a business replaces workers with AI, where do the saved wages go? Either lower prices or go straight to the bottom line. In both cases, the state can capture a part of the windfall.
Price drops can be snagged via an equal-and-opposite increase in VAT. This keeps shelf prices stable, with the proceeds channelled to raising the income floor.
Rising profits are trickier; hiking tax rates drives capital away. But if AI allows profits to soar, tax receipts will jump even without changing rates. So map pre-AI corporate and capital gains tax trends, and as revenues exceed that baseline, ringfence 75% of the excess for the income floor. Do the same with the government's own AI-driven savings.
These mechanisms are powerful yet imperceptible: prices unchanged for consumers, tax rates unchanged on earnings. A simulation at mosaicmodel.org shows that this alone could guarantee a lower middle-class income for all, ending poverty as we know it. Add more visible mechanisms, windfall tax, AI tax, land tax, and that floor rises to the upper middle class. Universal high income incarnate.
The math works, and the path is clear. What's missing is political will. If we act decisively, AI can end poverty and enable prosperity for all. If not, inequality will deepen, and faster-moving nations will leave us behind. The future AI pioneers dream of is within reach, but only if we reach for it.