‘We are driving in the fog’: Hundreds of economists admit they’re flying blind on AI

That alarmed voice is coming straight from the top of the economics profession. That’s Anton Korinek, a University of Virginia economics professor and one of the organizers of a statement signed this week by over 200 economists—including 16 Nobel laureates and the chief economists of OpenAI and Anthropic—admitting, in effect, that the profession is flying blind on AI.

The short statement, “We Must Act Now,” released Monday, doesn’t claim to have answers and claims the field is dangerously behind on the questions. It only offers three warnings.

1. AI may become radically more powerful over the next 10 years.
2. This could drive an unprecedented transformation of our economy, larger than the Industrial Revolution, but unfolding over a vastly shorter time frame. It could bring risks, including large-scale job displacement, as well as opportunities such as major gains in living standards.
3. Economists, policymakers and technology leaders must 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.

Michael Spence, a Nobel laureate at NYU, called for an “all hands on deck” approach to “steering AI in beneficial directions”—given the scale, scope and speed of its advancements, along with the high degree of uncertainty about how big and when the impact will arrive.

That cuts to the heart of this document’s unusual nature: it’s not a policy platform or a set of predictions. It’s a demand for the infrastructure to see clearly, issued by people admitting they currently can’t. One top Wall Street analyst, who did not sign the statement, thinks he knows what the confusion is all about.

Even the skeptics say they’re guessing

What makes the fog metaphor credible rather than alarmist is who’s admitting it. Nela Richardson, ADP’s chief economist, has described much of the public debate over AI and employment comes down to “guesswork” given the huge range of variables involved.

Likewise, Daron Acemoglu, the MIT Nobel laureate who has been the field’s most rigorous skeptic of AI productivity claims, previously told Fortune that he finds much of the productivity discourse around AI “brainless” even as he now says recent advances have made him more worried about near-term disruption. Signing onto today’s statement, Acemoglu said he was “so happy to join other leading experts in calling for the urgent need to redirect AI so that its risks are minimized and it can work for the benefit of workers and society.”

Stanford’s Erik Brynjolfsson, who The Atlantic recently noted has a remarkable knack for getting people who disagree to work together, previously told Fortune that he was puzzled by many of Acemoglu’s projections on AI and productivity. It’s no surprise, then, that Acemoglu signed on and that Brynjolfsson was an organizer of this statement.

“AI capabilities are advancing far faster than our understanding of the economic implications,” Brynjolfsson said in today’s statement. “We must act now to guide AI to complement humans rather than simply imitate them—and to generate prosperity for the many, not just the few.”

Brynjolfsson is at work trying to build new tools to understand the impact: his Canaries Dashboard with ADP Research tracks 4.6 million workers across more than 730 occupations in near-real time and purports to show employment for workers ages 22 to 25 in AI-exposed occupations shrinking more than 4% annually, even as the aggregate labor market looks calm. Named after the proverbial canary in the coal mine, Brynjolfsson previously talked to Fortune about why “flying blind” is such an urgent problem: the danger is invisible at the headline level and only shows up once you cut the data by age and task exposure. “We are flying blind into one of the most consequential periods in world history,” he said at the dashboard’s launch. “We need timely, trusted evidence to understand where AI is creating value and where it is disrupting work.”

The instruments themselves are contested

Torsten Slok, Apollo Global Management’s influential chief economist, published a blog post a week ahead of the “We Must Act Now” statement, arguing that even the basic concept of “AI exposure”—the term at the center of most labor market research—is contested by five competing measurement frameworks, each of which produces different results.

One measures what workers actually do with Claude, using real chat logs. Another does the same with Microsoft Copilot. A third has human experts judge which job skills AI is theoretically capable of replacing, regardless of whether anyone is using it that way. A fourth asks ChatGPT to grade its own usefulness on each task, and a fifth scans employer job postings for AI skill mentions. The theoretical frameworks run systematically higher than the usage-based ones because they ignore whether adoption is actually happening or worth the cost.

The disagreement is worst exactly where the stakes are highest. “What is most striking,” Slok writes, “is that the five measures disagree most exactly where the stakes are highest, among the very jobs everyone wants to flag as at-risk—such as telemarketers, tax preparers and writers.” There is a big difference, in other words, between asking “What could AI do to this job?” and asking “What are workers actually using AI for?” The theoretical use cases (questions 3, 4 and 5) are running “systematically higher” than the real-world ones (1 and 2), he said, because they ignore whether adoption is even happening or worth the cost.

As he often does, Slok offered a “bottom line” takeaway: “When someone says a job is ‘highly exposed to AI,’ the honest first question is: Exposed by which measure, and measuring what? Until that is pinned down, the label ‘AI exposure’ carries far less meaning than it appears to.”