Not yet human-reviewed
This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.
Economic Scenarios for Transformative AI
Citation. Korinek, A., Jones, C. I., Sacher, S., Cotter, T., & McCrory, P. (2026). Economic Scenarios for Transformative AI (The Anthropic Institute Working Paper No. 2026-02).
Summary. Korinek et al. present a task-based model that maps assumptions about AI capability, diffusion, per-task productivity, automation versus augmentation, and labor-market adjustment into U.S. economic outcomes through 2030. The authors use the framework to compare modest, substantial, and extreme change scenarios rather than to assign probabilities or make point forecasts. The paper also surveys 10,980 U.S. adults, whose median assumptions imply outcomes close to the authors’ substantial-change scenario.
Key claims
- In the model’s modest scenario, GDP in 2030 is 1.6% above a no-AI counterfactual; in the substantial and extreme scenarios, it is 8.3% and 32.4% higher, respectively.
- The extreme scenario combines rapid automation of cognitive work with transition frictions: annual GDP growth reaches 15%, the labor share falls from 60% to 45%, and nearly one in five cognitive workers is unemployed by 2030.
- Across the scenarios, AI-driven reallocation lowers relative demand and wages in cognitive occupations while increasing demand and wages in other occupations; the resulting unemployment depends on the difficulty of changing occupations.
- The model estimates that automating research produces relatively small additional productivity growth by 2030 because physical-task bottlenecks remain in its semi-endogenous growth framework.
Evidence & limitations
- The analysis is a calibrated scenario model informed by task-based production assumptions and a representative survey of 10,980 U.S. adults, not causal evidence or a probability forecast.
- The model simplifies worker heterogeneity and capital, and omits catastrophic risks, political economy, business cycles, financial disruptions, price-rigidity demand feedback, household production, and rapid advances in robotics.
- Public provenance was verified against Anthropic’s public PDF, which has the same SHA-256 as the ingested Drive artifact.