Signals: Geoff Hinton Not Wrong After All?

One data point: A few days ago (3/25/26) Mitchell Katz, CEO of NYC Health + Hospitals — the largest public hospital system in the US — told a Crain's panel he's ready to replace radiologists with AI for first reads, pending regulatory changes. David Lubarsky, CEO of Westchester Medical Center Health Network, said their AI mammography tool misses only 3 in 10,000 negatives. Sandra Scott, CEO of a small Brooklyn hospital system, called it "a game-changer."

Geoffrey Hinton

One implication: Radiologists have spent nine years celebrating the "failure" of Nobel prizewinning AI scientist Geoffrey Hinton's famous 2016 prediction that we should stop training radiologists because we already have enough to last until the machines take over. Now the CEO of America's largest public hospital system says he's ready to let the machines take over — at least for first image readings. Katz isn't talking about firing radiologists tomorrow. He's talking about letting AI handle first reads and having radiologists review the flags. And he's not a researcher, he's a hospital CEO trying to solve a staffing and cost problem, asking his peers on a public stage why they aren't pushing regulators to let it happen.

The radiology community's response has been predictable: another confidently uninformed administrator being duped by AI hype. Maybe. But Katz runs eleven hospitals. Scott runs a small system where every dollar matters. These aren't people chasing novelty. They're people doing math.

Hinton recently acknowledged he was wrong on timing. But the endpoint he suggested — machines taking over image reading — just got boosted by three hospital CEOs on a public stage.