The issue is that workers have almost no control over when their economic gains count as income, while owners of immense fortunes have far more control over when—and sometimes whether—their gains become taxable.
That isn’t just a loophole. It is a tax system designed for an earlier economy, when most income arrived through wages and billion-dollar personal fortunes were far less common.
A practical reform doesn’t require the IRS to appraise every private company, painting and yacht each December. Rates can rise on extraordinary incomes and extraordinary capital gains. Publicly traded billionaire assets can be taxed when market prices establish the gain. Private gains can be taxed when they become measurable and usable—through a sale, transfer, enormous personal cash-out or inheritance.
At the same time, ordinary homes, retirement accounts, farms, small businesses and genuine one-time sellers should be protected. Someone selling a business built over 30 years shouldn’t be treated like someone receiving millions every year.⁵
The mechanics belong in legislation. The principle is easier to understand:
Teachers pay as income arrives. Billionaires should pay when their gains are priced, cashed out, transferred or inherited.
But fixing the tax code is only preparation for the much larger change ahead.
Artificial intelligence may prove as economically transformative as the Industrial Revolution, and it may unfold much faster. It is already moving beyond repetitive clerical work. AI can write software, analyze medical images, prepare legal research, translate languages, design products, tutor students and help scientists identify promising drugs.
It will make mistakes. It will need supervision. Many occupations will be changed rather than eliminated. But “transformation” is a reassuring word for a process that can still eliminate positions, reduce hours and weaken wages.
AI doesn’t have to replace an entire profession to reduce the number of people employed in it. If ten workers using AI can produce what once required twenty, the company may employ ten people—or keep twenty and produce twice as much.
The technology itself doesn’t decide who receives the gain.
Ownership, bargaining power, competition, taxation and public policy do.
The IMF estimates that AI could affect almost 40 percent of employment worldwide. In advanced economies, where more jobs involve cognitive and administrative work, exposure may be substantially greater. It also warns that AI could deepen inequality by raising returns to capital and disproportionately rewarding people who already own substantial assets.⁶ More recent IMF research reaches an even sharper conclusion: AI may increase wealth inequality as capital owners capture a large share of the gains.⁷
That is the central economic question of the AI age.
For generations, employment has been our principal distribution system. Most people worked, employers paid them, and wages gave them a claim on the prosperity the economy produced.
AI may weaken that connection.
It could allow society to produce far more medicine, transportation, education, entertainment and useful information with far less human labor. That could become one of the greatest advances in human history. We might work fewer hours, retire earlier, care for our families better and spend more of our lives doing things machines can’t do for us.
Or AI could create an economy in which a relatively small ownership class controls the machines while everyone else competes for the work that remains.
The danger isn’t that AI will make us poor. It is that AI will make society spectacularly rich while much of the population becomes economically unnecessary.
Productivity is not a distribution system.