AI · Studies · 2026-09-11 05:04:00
An experiment pitted five diverse AI agents against five clones over 14 nights, yielding 4,130 scored forecasts. The diverse lenses stayed apart, but a base rate the author got 15x wrong tied both. The study questions whether agent diversity improves forecasting, pointing to base-rate calibration as decisive.
Auto-generated summary for reference only. RSS