.png)
1 month ago
6
Working Paper 34312
DOI 10.3386/w34312
Issue Date October 2025
How much can AI accelerate progress in different research fields? This paper shows that three features—the share of research tasks AI performs, the productivity of AI at those tasks, and the strength of bottlenecks—are key determinants of AI’s implications in any area, from cancer therapeutics to software design. The model maps changes in AI capabilities to research outcomes, quantifies the “marginal returns to intelligence,” and shows how AI can shift returns to R&D investment. Concepts like superintelligence, Powerful AI, and Transformative AI are further engaged and disciplined. Finally, the framework sets a measurement agenda linking AI benchmarks to field-specific opportunities for accelerating progress.
More from the NBER
- Feldstein Lecture
- Presenter: N. Gregory Mankiw
N. Gregory Mankiw, Robert M. Beren Professor of Economics at Harvard University, presented the 2025 Martin Feldstein...
- Methods Lectures
- Presenters: Raj Chetty & Kosuke Imai
SlidesBackground materials on mediationImai, Kosuke, Dustin Tingley, and Teppei Yamamoto. (2013). “Experimental Designs...
- Panel Discussion
- Presenters: Oleg Itskhoki, Paul R. Krugman & Linda Tesar
Supported by the Alfred P. Sloan Foundation grant #G-2023-19633, the Lynde and Harry Bradley Foundation grant #20251294...
