Public forecasting contests (prize pools and curated questions) are becoming institutional benchmarks that measure AI versus human forecasting skill in the wild. Because questions are sourced, resolvable, and monetarily incentivized, these contests produce reproducible, policy‑relevant data on model calibration, timeliness, and areas where AI adds or undermines collective judgment.
— If prediction contests become standardized benchmarks, they will shape public trust in AI, influence funding and regulation of forecasting systems, and determine whether algorithmic forecasts are used in policy and markets.
Scott Alexander
2026.11.21
100% relevant
Metaculus’ 2027 ACX contest will allow AI forecasting agents, will post AI forecasts on every question, and highlights AI now competing with strong human forecasters (article quote: "the best are now competitive with strong human forecasters").
← Back to all ideas