New paper maps AI benchmark scores to AI exposure across 894 occupations, 2020–2026
krisgulati · x · 2026-10-13
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. Researchers from the Felten/AIOE line engaged, noting its advantage of continuous updates as benchmarks evolve.
More from Research
- DatologyAI launches Datology Labs for open collaborative data science research — pratyushmaini · 2026-10-15
- Prime Intellect experiments show ECHO beats pure RL in turn-capped, no-thinking robotics-like environments — DimitrisPapail · 2026-10-15
- Incoming Berkeley professor lists open questions: pre/post-training alignment, safe agents — canondetortugas · 2026-10-15
- Berkeley's QF3: off-policy flow RL with filtered Q-gradients trains humanoids from scratch — berkeley_ai · 2026-10-15
- VLMs May Turn Clearly Superhuman in Vision Within a Year or Two, Researcher Predicts — arnabdotorg · 2026-10-15
- Andrew Perrault: eval environments released only when scorers match human majority — arnabdotorg · 2026-10-15