Radiologists up 17% since 2016 despite AI tools

What happened

The article explores why Geoffrey Hinton's 2016 prediction that radiologists would be obsolete was wrong. Despite the FDA approving more than 1,000 AI radiology tools, the number of radiologists has risen by 17 percent since 2016, vacancy rates are near all-time highs, and average salary has increased from about $350,000 to $570,000. The piece introduces three questions to predict whether a job faces replacement risk.

Why it matters

A job's vulnerability to AI depends on whether its clean tasks (predictable, data-heavy work) can be separated from messy tasks (unpredictable, relationship-heavy work) without harming performance. Trial lawyers need to do their own case-prep to cross-examine effectively in court; recruiters and software engineers can offload routine work and still do their core job well. For weak-bundle jobs, automation has historically triggered higher employment—not lower—when cheaper services make people want more of them, as happened with automobiles after the assembly line, ATMs, and spreadsheets.

What to watch

Job openings for recruiters rose 30 percent from 2023 to 2025; for software engineers, they have doubled. Call-center worker numbers are rising fast even as firms use AI to handle customer-service requests. The critical factor remains whether demand for the service grows when it becomes cheaper—food work fell from 40 percent of jobs to 1 percent because people can eat only so much, whereas accounting clerk roles collapsed while accounting specialist roles grew higher-wage.