Economist Declines AI 'Act Now' Statement, Citing Vague Policy Proposals

The author received a request to sign a statement titled "We Must Act Now: A Statement on AI's Transformation of the Economy," which has gathered signatures from many prominent economists. The statement argues that AI may become radically more powerful over the next 10 years, could drive an unprecedented economic transformation larger than the Industrial Revolution but over a vastly shorter time frame, and could bring both risks—including large-scale job displacement—and opportunities such as major gains in living standards. It calls for economists, policymakers, and technology leaders to "act now to understand the economics of transformative AI and to build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society."

The author declined to sign because the statement provides no concrete policy proposals. He worried that signing it would effectively commit him to an unknown second statement with specific policies that the authors would eventually release. More critically, he recognized that the statement's language about "steering AI in a direction that complements humans" echoes economist Daron Acemoglu's ideas from Power and Progress. Acemoglu himself signed the statement, and revealed that he had revised the initial version before agreeing to sign, specifically to secure the language about "steering AI" in a direction that complements humans. This, Acemoglu wrote, reflects "what I have been arguing for over a decade now," and he argued that "good AI needs to complement humans" and that "the current focus on AGI is, in all but name, an agenda for displacing humans from meaningful work."

The author argues that the idea of "steering" AI is flawed on two fronts. First, it is essentially impossible: inventors cannot predict how technologies will ultimately be used by businesses. James Watt, inventing the steam engine in 1765, had no way to predict most of its applications or whether it would ultimately create or destroy jobs. This historical pattern repeats with AI: current employment data for people aged 20–24, 25–54, and young college graduates show rates "just about the same as they were before ChatGPT ever existed," and one recent study found that companies adopting AI hire more workers than competitors in the same industry. Geoffrey Hinton, a co-inventor of modern AI, famously predicted the end of human radiologists within a few years, only to see radiologist hiring and salaries boom as AI complemented rather than replaced their work.

Second, the author questions who would do the "steering." Acemoglu's own book does not answer this question convincingly, and the author argues that assembling a panel of experts to assess the labor market effects of technology that has not yet been invented is "probably an impossible task." Such a panel would be "operating purely on gut instinct and quasi-science-fictional supposition," introducing randomness and uncertainty into the innovation process. The author points out that Acemoglu himself has a poor track record on prediction: his empirical paper claiming robot-buying companies hire fewer workers was contradicted by many follow-up studies, and his theoretical paper on AI and productivity was built on faulty assumptions.

Given that employment data currently show no visible wave of AI job destruction, the author concludes that inaction may be the right policy. Historical precedent—collectivization of agriculture, Mao's "backyard production"—shows that heavy-handed intervention in entire sectors can "crash and burn spectacularly." It is possible, he argues, that "the AI we're building now is already highly complementary to human workers, and that the best approach is not to 'Act Now', but to simply sit there and do nothing." The author remains open to signing the statement if its authors provide concrete policy specifics, but for now it seems to hide "genuinely inadvisable Acemoglu-ism behind a screen of extreme vagueness."

The author is responding to a growing coalition of economists who have signed a statement calling for immediate action on AI's economic effects, a statement that appears to have been shaped through negotiation with Daron Acemoglu, a prominent economist and author of Power and Progress. Acemoglu himself revealed that he had "a hand in revising" the statement after the organizers reached out, and he credits his own decade-long argument for AI that complements rather than displaces humans as his reason for signing.

The author's core objection is not to action on AI per se, but to the claim that policymakers and technologists can meaningfully "steer" AI development toward beneficial labor market outcomes. He supports this skepticism with two categories of evidence: historical observation (inventors routinely fail to predict how their creations will be used) and recent data (current employment metrics show no visible AI-driven job destruction, contrary to widespread predictions). The author also cites Acemoglu's own poor track record on prediction—a paper claiming robot buyers hire fewer workers was contradicted by many follow-up studies, and a theoretical paper claiming AI would not raise productivity was based on "arbitrarily assuming away parts of [Acemoglu's] own model." This sets up the author's reductio: if technologists and economists themselves cannot predict labor market effects, how can a panel of experts credibly "steer" innovation toward desired outcomes? The author argues such a panel would add "noise" to the innovation process and be "simply adding noise to the innovation process, offering rewards and punishments essentially at random."