AI & ML

Reliable, and still wrong

A large-scale audit of AI-as-judge evaluation — covering over half a million individual judgments — finds that AI judges are consistently reliable b…