The Upside Is Concentrated. The Downside Is Not.
Joe Stiglitz, the Columbia University economist who won the Nobel Prize for his work on information asymmetries, is not opposed to technological progress. He is opposed to the assumption that progress distributes itself fairly.
In a recent interview with Fortune, Stiglitz argued that artificial intelligence will follow the same pattern as prior waves of automation: the people who own the technology get richer; the people whose labor it replaces get less. What makes this moment different, he says, is the speed of displacement and the political posture of the industry driving it.
The Contradiction at the Center
"Unfortunately, the tech bros, who are obviously advocates of this, are at the same time pushing for smaller government," Stiglitz told Fortune.
That sentence carries a lot of weight. The same executives and investors accelerating AI deployment are, in many cases, funding political efforts to reduce the size and scope of federal programs — unemployment insurance, job retraining, social insurance — that would be the primary buffers for workers displaced by the technology they are building.
This is not a coincidence. It is an incentive structure. If you capture the gains from AI and bear none of the adjustment costs, smaller government is a rational preference. The workers absorbing those costs have a different set of interests.
What the Labor Market Math Looks Like
Stiglitz's concern is grounded in how productivity gains have historically been distributed. When technology raises output per worker, the question is always who captures that surplus — shareholders, executives, or the broader workforce. The evidence from the last four decades of automation suggests the answer has consistently favored capital over labor.
AI, in Stiglitz's view, accelerates that dynamic rather than reversing it. The technology is particularly good at tasks that were previously the domain of knowledge workers — the segment of the workforce that had, until recently, been relatively insulated from automation pressure.
The Governance Gap
The policy implication Stiglitz draws is direct: if AI's gains are going to be broadly shared, government has to be large enough and capable enough to do the redistribution. That means progressive taxation on AI-driven profits, investment in public education and retraining, and social insurance robust enough to handle structural unemployment — not cyclical unemployment.
None of that is compatible with the deregulatory, small-government agenda that significant parts of the tech industry have backed politically.
What Executives Should Take From This
For business leaders, Stiglitz's argument is not just a political science lecture. Companies deploying AI at scale are making workforce decisions that will affect communities, tax bases, and consumer purchasing power. The macro risk — a workforce that cannot afford to buy what AI-enabled companies produce — is a business risk, not just a social one.
Leaders who treat AI deployment as a pure cost-optimization exercise, without accounting for the downstream labor market effects, are externalizing costs that will eventually come back in some form: regulatory, reputational, or economic.
Stiglitz has spent his career documenting what happens when markets are allowed to concentrate gains without mechanisms for redistribution. His read on AI is that the mechanism is missing — and the people best positioned to build it are actively working to prevent it.