How Accurate Have AI Progress Forecasts Been So Far?

(forecastingresearch.substack.com)

2 points | by gmays 13 hours ago ago

2 comments

  • michaelastreiko 12 hours ago ago

    The calibration angle is useful: separating capability forecasts from deployment and adoption forecasts makes misses easier to diagnose.

  • katamin 12 hours ago ago

    I think this will widen the gap even more.