Digging into some of Anthropic’s research on the early effects of AI and employment, one thing that jumped out to me: Two professions that had about the lowest ratio of observed to theoretical AI use are “Healthcare Practitioners” and “Life and Social Sciences.” Some of this may relate to regulations/legal risks (especially healthcare), but I would argue a lot of this is because biology and medicine is inherently much messier than coding or math, and less well-suited to probabilistic pattern-matching.