New paper maps AI benchmark scores to AI exposure across 894 occupations, 2020–2026
robseamans · x · 2026-10-14
Kris Gulati and Jake Prokopets released a preliminary paper asking whether AI benchmarks can predict which jobs AI will disrupt.
Method pipeline (per figure):
- Collect AI benchmarks (FrontierMath, SWE-bench, MMLU, MMMU, etc.)
- Map each benchmark to ONET human abilities (mathematical reasoning, written comprehension, problem sensitivity...)
- Assess economic relevance: human-comparative capability and workplace transferability
- Ability exposure = capability × transferability, aggregated to occupations via ONET importance weights
It links AI progress to 894 occupations from 2020–2026 and asks whether benchmark results signal where AI adoption emerges next. Commenters note similarity to Felten et al.'s AIOE index, with the advantage of continuous updates as benchmarks evolve.
More from AGI Musings
- Gary Marcus interview eases AI panic concerns, but dodges Hinton-style doomsday warnings — Ok_Pangolin_9134 · 2026-10-14
- AI-native game worlds in minutes: 'dinosaurs' still bolting AI onto game engines, argues dev — nptacek · 2026-10-14
- Fchollet: Autoregressive Sampling Alone Can't Solve Real-World Robotics — fchollet · 2026-10-14
- Zvi: Safety culture at frontier AI labs is broken — even Anthropic falls short — TheZvi · 2026-10-14
- Oxford, Cambridge, ETH Hiring Shifts: Papers Count Less, Understanding Is the Rare Skill — littmath · 2026-10-14
- Chollet: Intelligence Needs Grounding — It's Almost Never the Bottleneck for Real Problems — fchollet · 2026-10-14