Yann LeCun has previously said that Dario Amodei’s idea of a “country of geniuses in a datacenter” is “bs”, and now he appears to feel similarly about Amodei’s claims of AI-related job losses.
In a post that has been making the rounds on X, the Turing Award winner and former Meta Chief AI Scientist delivered a blunt verdict on Amodei’s recent warnings about AI and employment: “Dario is wrong. He knows absolutely nothing about the effects of technological revolutions on the labor market.”
The rebuke is directed at claims Amodei made in a series of high-profile interviews last month, warning that AI could eliminate roughly 50% of entry-level white-collar jobs within five years and push unemployment to 10–20%. Speaking to Axios, Amodei said the shift could happen “almost overnight” as companies quietly stop hiring and replace human workers with AI agents — particularly in fields like law, finance, tech, and consulting.
A Pattern of Pushback
This is not the first time LeCun has publicly taken issue with Amodei’s vision of the AI future. Earlier this year, he called the “country of geniuses in a datacenter” idea “complete BS”, arguing that current large language models cannot scale to human-level intelligence. He has also previously called Amodei “deluded” about the dangers of AI and accused Anthropic of pursuing regulatory capture by stoking fear with what he characterised as dubious studies.
LeCun’s latest post, however, is notable for a different reason: he is not simply defending his own technical views. He is explicitly directing people away from listening to AI lab leaders — including himself — on questions of economic impact. “Don’t listen to him, Sam, Yoshua, Geoff, or me on this topic,” he wrote, naming Amodei, Sam Altman, Yoshua Bengio, and Geoffrey Hinton alongside himself.
Instead, he pointed to economists who have spent their careers studying how technological change reshapes labour markets: Philippe Aghion, Erik Brynjolfsson, Daron Acemoglu, Andrew McAfee, and David Autor.
AI Leaders vs. Labour Economists
The distinction matters. Amodei’s warnings have been framed as insider candour — a rare admission from a frontier AI CEO willing to publicly voice concerns that others only whisper. He has acknowledged the contradiction of warning about a technology he is simultaneously building and selling, but argued that transparency is a moral obligation.
LeCun’s argues that building AI does not make someone qualified to forecast macroeconomic effects on employment. The history of technological disruption is a specialised field, and it is one where the economists LeCun names have built entire bodies of research.
Acemoglu, for instance, has published extensively on whether automation creates or destroys jobs in net terms — and his conclusions are considerably more nuanced than blanket displacement forecasts. Brynjolfsson and McAfee’s work on the “second machine age” argued that technology tends to restructure labour demand rather than simply eliminate it. Autor has documented how automation hollows out middle-skill work while creating demand at both the high and low ends.
None of these economists have endorsed a scenario as stark as Amodei’s 50% figure.
The Incentive Problem
There is a harder question lurking in LeCun’s dismissal, one he has gestured at before: whether AI executives making apocalyptic labour-market predictions have incentives that should make us sceptical. David Sacks has also hinted at something similar — he’s said that Anthropic uses ‘fear’ as a marketing tactic.
Predictions of mass disruption can serve multiple purposes simultaneously — they can be sincere warnings, they can attract investment, and they can justify calls for regulation that, in practice, favour well-resourced incumbents over smaller competitors. LeCun has previously argued that Anthropic’s safety-focused communications are a deliberate strategy to push for rules that would benefit larger labs.
Amodei, for his part, has proposed a “token tax” on AI usage to help manage the transition and has called on lawmakers to act. Whether that constitutes responsible stewardship or self-interested positioning depends heavily on who you ask.
What Economists Actually Say
The economists LeCun cited do not uniformly agree with each other, which is itself part of the point. Acemoglu has been among the more pessimistic in recent years, warning that AI investment has been heavily skewed toward automation rather than productivity-enhancing tools that complement human workers. Brynjolfsson has been more optimistic, arguing that the productivity gains from AI will eventually translate into broader prosperity, though the transition may be painful.
What they share is a framework built on historical evidence — the mechanisation of agriculture, the electrification of factories, the rise of computing — rather than extrapolations from current model capabilities. Technological transitions have consistently created new categories of work even as they destroyed old ones, though the timelines and distribution of those gains have been deeply uneven.
None of that makes Amodei definitively wrong. But it does suggest that a 50% headline figure, delivered by the CEO of one of the companies driving the disruption, deserves more scrutiny than it has received.