A study by the Forecasting Research Institute shows that leading AI experts systematically underestimated the pace of industry development. According to data recounted by The Decoder, AI achieved a gold medal level at the International Mathematical Olympiad five years earlier than the experts' median forecast.
Estimates also diverged significantly from the dynamics of Anthropic's annual revenue: the actual figure was approximately five times higher than expert forecasts. Meanwhile, forecasts for practical applications, including autonomous vehicles, presented a more ambiguous picture.
This is important for companies, regulators, and research groups that use expert assessments for planning. A single metric is insufficient: the speed of progress varies by task, and available information is presented in a brief recounting of a single source.
editorial commentary
Why it matters
A likely consequence is a revision of AI forecasting methods, separating assessments by individual tasks rather than by industry as a whole. The next observable signal will be the publication of the study's primary materials and new comparisons of forecasts with actual results. Significant uncertainty remains because currently only a retelling of a single source is available.