Brief
Pilot study asks experts to forecast AI R&D acceleration impacts
METR and the Forecasting Research Institute ran a pilot survey of 8 AI experts and 10 superforecasters. The survey asked them to estimate the likelihood of rapid acceleration and transformative impacts conditional on hypothetical evaluation results.
METR and the Forecasting Research Institute conducted a pilot survey of 8 AI forecasting experts and 10 superforecasters. The survey asked participants to estimate the likelihood of rapid acceleration from automated AI R&D and transformative impacts, conditional on hypothetical evaluation results. The study aims to clarify how to interpret AI R&D evaluation results and where experts disagree. As a pilot, it is an initial exploration, not a definitive forecast.
Our reading
Our reading is that this pilot study is a first step to map expert disagreement, but its small sample and hypothetical scenarios mean it cannot yet tell us how likely rapid acceleration is.
What to do or watch
Readers should not treat this pilot as a forecast of rapid acceleration, since its small sample and hypothetical scenarios leave the likelihood unresolved. The key question to watch is how experts interpret actual AI R&D evaluation results and where their disagreements lie.
Source details and supporting facts
Each line is stated by the page named above it.
Stated by metr.org
- AI agents are improving rapidly at autonomous software development and machine learning tasks.
- If recent trends hold, AI agents may match human researchers at challenging months-long research projects in under a decade.
- Some economic models predict that automation of AI research by AI agents could increase the pace of further progress dramatically.
- Significant disagreement remains in interpreting the implications of AI R&D measurements.
- METR worked with the Forecasting Research Institute to conduct an initial pilot survey of 8 AI forecasting domain experts and 10 superforecasters.
Sources
- METRText stored 15 September 2026
How this story was checked. Written from the 1 page listed above, stored 15 September 2026; claims checked against that stored text on 15 September 2026.
What that means
- 5 of 5 reported statements were confirmed against the page that carries them; the rest were removed rather than published.
- Figures in the text were required to appear in the stored source text: yes. Identifiers: yes.
- The check reads stored text only: no claim rests on a fresh look that did not happen.
- Where the reporting was silent, the text says so instead of filling the gap.