New paper uses 288M job postings to measure where AI raises the value of work it can't do
soumitrashukla9 · x · 2026-10-22
Soumitra Shukla released a new paper, "Technology Bottlenecks: Measuring Where AI Could Raise the Value of Work It Cannot Yet Do."
Key points:
- Unlike AI exposure measures that ask what AI can do, this asks where the marginal value of AI-infeasible work rises as AI speeds up the rest of the job
- Method: 288 million U.S. job postings mapped line by line to their closest ONET tasks, combined with GPT-4 feasibility ratings from Eloundou et al. (2024), to build a "bottleneck pressure" measure per task per occupation
- In a task-based framework, for an AI-infeasible task the pressure relates to its exposure
Daniel Rock praised it: bottlenecks are the story when it comes to local productivity gains.
Related event: Paper Uses 288M Job Postings to Measure AI's "Bottleneck Pressure"(4 posts)→
More from AGI Musings
- Survey: 54% of Workers Avoid Company AI Tools, Waste 8 Hours a Week Fixing AI Mistakes — marigo · 2026-10-22
- ML pioneer Dietterich: OpenAI's agents ran for months with essentially no supervision — tdietterich · 2026-10-22
- Altman hints OpenAI's internal models do math public models can't, flags RSI speed limits — VraserX · 2026-10-22
- New symmetric Venn diagrams with 17, 19 and 23 curves found via AI-assisted search, breaking 2014 record — pickover · 2026-10-22
- Ex-OpenAI researcher's hot take: NYC trails SF by 4-7 months in AI, London 12+ — jefrankle · 2026-10-22
- Debate on AI proof formalization: mass-formalize textbooks first, then let AI build on the repos — jacobaustin132 · 2026-10-22