Three Ways to Assess AI's Threat to Jobs
The Atlantic argues that asking "Can AI do my job?" is the wrong frame and proposes three alternative questions to better assess occupational risk, per its June 11, 2026 feature. The article cites a 2016 remark from Geoffrey Hinton, which it quotes as saying, "people should stop training radiologists now," and contrasts that remark with sector outcomes. According to The Atlantic, since 2016 the number of radiologists has risen by 17 percent, vacancy rates are near all-time highs, and average salaries increased from about $350,000 to $570,000, making radiology the third-highest-paid U.S. medical speciality. The piece also quotes Dario Amodei, who last year said AI would "wipe out half of all entry-level white-collar jobs," as an illustration of prevailing fears. Editorial analysis: For practitioners, the example highlights that automation can expand demand by increasing throughput and creating new work rather than only substituting labor.
What happened
The Atlantic published a feature titled "Three Ways to Think About AI and Jobs" on June 11, 2026. The article reproduces a 2016 remark from Geoffrey Hinton, quoting him as saying, "people should stop training radiologists now," and juxtaposes that statement with recent labor outcomes in radiology. According to The Atlantic, since 2016 the number of radiologists has risen by 17 percent, vacancy rates are near all-time highs, and average salaries increased from about $350,000 to $570,000, making radiology the third-highest-paid medical speciality in the United States. The article also cites a recent claim by Dario Amodei, who the piece quotes as saying AI would "wipe out half of all entry-level white-collar jobs."
Editorial analysis - technical context
The Atlantic frames the central mistake as focusing narrowly on whether AI can replicate task-level performance. Industry-pattern observations: automation often changes throughput, task composition, and demand for complementary human skills. In fields where AI accelerates diagnosis or processing, institutions can expand service volumes, which may increase headcount or shift roles toward interpretation, oversight, and procedures that require human judgment.
Context and significance
Editorial analysis: The radiology example illustrates a broader labor-market dynamic where supply, demand, and institutional incentives interact with automation. Observed patterns in comparable transitions show that certification, legal liability, workflow integration, and reimbursement structures materially affect whether AI substitutes for or augments human labor. These systemic factors can outweigh raw model capability when predicting occupational outcomes.
What to watch
Editorial analysis: Observers should track reimbursement and regulatory changes, vacancy and hiring trends by specialty, and the decomposition of work into automatable versus nonautomatable tasks. For practitioners evaluating career risk or workforce planning, monitoring how AI deployment alters service volume, billing incentives, and required human oversight will be more informative than benchmark-level comparisons of model accuracy alone.
Scoring Rationale
The piece reframes a common public question about AI and jobs into operationally useful diagnostics for practitioners and policy observers. It is notable for labor-market implications rather than technical novelty, so it rates as a mid-level story important for workforce planning.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
