WrongPrediction
Median forecast for the maximum compute used in an AI experiment by 2024 was about 6x too low
“45. Maximum Compute Used in an AI Experiment 100,000 578,703.7 1.92 33”
Context: Existential Risk Persuasion Tournament (XPT), June-Oct 2022; group median forecasts as reported by the Forecasting Research Institute in "Assessing Near-Term Accuracy in the XPT" (2025). Units as in the XPT question (resolution 578,703.7 vs median 100,000).
How we judge it
Our interpretation: treated as a point forecast; correct if within a factor of 2 of the resolved value.
Resolution criteria are our interpretation of the claim, written before grading. Methodology
- Said
- Resolve by
- Verdict
- Wrong ()
- Stated probability
- None stated
- Specificity
- 3 of 3 (precise and measurable)
- Score
- 0.0 points × weight 3
Evidence & grader’s note
- FRI near-term XPT accuracy report, Table 3.1: resolved at 578,703.7 (superforecasters' largest miss)
The superforecasters' largest standardized error: actual maximum training compute was about 5.8x their median.
