We Need To Do Nothing About AI
Fear of new technology is hardly new.
In recent weeks, a letter on AI titled “We Must Act Now” has made headlines. Its list of signatories brought together an unlikely coalition of intellectuals, from Joseph Stiglitz and Paul Krugman to Alex Tabarrok, Tyler Cowen, and Niall Ferguson.
The signatories agree on three key propositions. First, they argue that “AI may become radically more powerful over the next decade.” Second, they argue that it could “transform the economy on a scale larger than the Industrial Revolution, but over a much shorter period, bringing both large-scale job displacement and significant improvements in living standards.” Third, they argue that “economists, policymakers, and technology leaders must act now to create the incentives, guardrails, and institutions needed to ensure AI benefits society.”
These propositions have created an unusual consensus among economists and policymakers. They also reflect a broader public anxiety about AI and its impact on people’s livelihoods. Yet, as Stanford economist John Cochrane remarked in response to the letter, “You see AI everywhere, except in productivity and labor statistics.” That is the first problem with the argument: those are the two places where the letter’s predictions should already be showing up.
The letter makes extraordinary claims, and as the famous saying goes, extraordinary claims require extraordinary evidence. Yet the evidence offered is remarkably thin. The language is dramatic—“large-scale job displacement,” “larger than the Industrial Revolution,” and “major gains in living standards”—but every prediction is qualified by one word: could. There is still little evidence that AI has fundamentally reshaped employment or productivity.
Take the labor market as an example. A widely cited survey claimed that employers plan to reduce graduate hiring because of AI. But the study was forecasting the future, not describing present reality. It found that “four in ten employers believe entry-level roles could disappear within the next five years.” Once again, the evidence rests on expectations and fears rather than observed outcomes.
Now compare that with what is actually happening. Economist Valentin Boboc has argued that entry-level jobs are growing fastest at firms that are aggressively adopting AI. If AI were already eliminating these positions, we would expect to see the opposite.
Consider youth unemployment. Which developed country currently faces one of the most severe youth employment crises? The answer is not a country overwhelmed by AI, but the United Kingdom. More than 1 million people aged 16–24 are classified as NEET (Not in Education, Employment, or Training), at an estimated annual cost of £125 billion ($169 billion) to the economy.
The greatest threat to young workers is not AI, but poor government policy. In the UK, higher employer National Insurance contributions have increased the cost of hiring. More generous long-term sickness benefits have weakened incentives to return to work for some claimants. Large increases in the minimum wage have also raised the cost of employing inexperienced workers. These policies provide a far more convincing explanation for weak youth employment than AI does.
The fundamental problem with the letter is its demand for urgent action in the absence of convincing evidence. It asks governments and institutions to intervene now to address a risk that has yet to materialize. The consensus itself appears to rest more on speculation than on observed facts.
As F.A. Hayek warned in his Nobel Prize lecture, quoting Alfred Marshall: “Students of social science must fear popular approval: evil is with them when all men speak well of them.” When there is overwhelming agreement that governments must “act now,” economists should become more skeptical, not less.
The mindset that governments must immediately regulate AI may prove more dangerous than AI itself. Markets generally adapt to technological change. When genuine market failures emerge, there is a case for government intervention. But where, exactly, is the market failure in AI today that demands immediate regulation?
Many of the economic problems we face today are the result of government failures rather than market failures. Excessive labor market regulation in many European countries has made hiring and firing employees so costly that employers become reluctant to take risks on new workers. Henry Hazlitt’s life contrasted this with the period before Franklin D. Roosevelt’s labor reforms:
In those years, it was not hard for a young man to get a job. With no government-imposed obstacles to hiring and firing, no minimum wage laws, no workday or workweek restrictions, and no unemployment or social security taxes, employer and potential employee needed only to agree on the terms of employment.
Fear of new technology is hardly new. Nor has it been confined to critics of free markets such as Stiglitz or Krugman. During the Industrial Revolution, David Ricardo added a chapter titled “On Machinery” to Principles of Political Economy and Taxation, arguing that machinery could often be “very injurious” to workers and might not always serve the general good. Technological pessimism has a long history even among champions of free markets.
That is why I am not surprised to see free-market economists among the signatories of this letter. What is surprising is their willingness to advocate immediate policy intervention without enough evidence that such intervention is needed.
Until there is clear evidence of widespread economic harm, we should do nothing about AI. Building regulations around what might happen, rather than what has happened, risks stifling innovation, protecting incumbent firms from competition, and erecting barriers to new entrants.
Mani Basharzad is a Research Associate at the Institute of Economic Affairs and an Asia Freedom Fellow at the London School of Economics. His work has been published by the New York Post, National Review, The Spectator, and Daily Express.
This piece was reproduced with the permission of FEE. Any opinions articulated herein are those of the author, not The Prickly Pear. To read the original piece, please visit HERE.
