Why Jensen Huang Called Geoffrey Hinton’s AI Warning “Irresponsible”

Nvidia CEO Jensen Huang’s objection to Geoffrey Hinton was about both evidence and consequences. In a September 23, 2026 interview on The Ezra Klein Show, Huang argued that Hinton’s estimate of a roughly 10–20% chance of catastrophic AI harm was not a scientifically established probability. He also called such predictions “hurtful,” pointing to an earlier forecast about radiologists as an example of how warnings can affect career choices. 25

Why Huang rejected the risk estimate

When Ezra Klein raised Hinton’s warning, Huang replied that the 10% figure was “not grounded on science” or research. “Just because it comes from a scientist doesn’t make it scientific,” he said. His point was not that AI safety is unimportant; it was that a precise-sounding number should not be mistaken for a measured probability simply because an expert offers it. Axios reported that Hinton himself had described estimates of this kind as a “wild guess.” 25

Why he brought up radiologists

In 2016, Hinton advised people to stop training as radiologists, predicting that AI would outperform humans at reading medical images within five years. Huang used the profession’s continued need for specialists, even as AI entered radiology, to challenge the leap from automating a task to eliminating a job. He argued that a student who took the warning literally might have avoided a career that remained viable. 214

That example helps explain “hurtful”: Huang believes confident forecasts can change people’s decisions before their accuracy is known. But a mistaken prediction about one profession cannot establish whether a separate warning about catastrophic AI is right or wrong. 214

The deeper split on safety, rules and work

Huang presents AI safety largely as an engineering responsibility: companies should not ship products they cannot control. He has resisted new AI-specific regulation while also saying AI firms should not receive exemptions from existing antitrust or liability law. Hinton’s warning puts greater emphasis on the possibility of catastrophic harm from increasingly capable systems. Neither position, nor the radiology example, supplies a reliable numerical estimate of that risk. 1714

On work, Huang’s radiology argument offers a useful distinction between jobs and the tasks within them. It is a reason to scrutinize predictions that AI will erase entire occupations—not a guarantee that every profession will adapt in the same way. 2314