Expert Roles in Model Training

Why experts writing AI evals is creating the fastest-growing companies in history | Brendan Foody

Topic 3 of 8

Published Sep 18, 2025 · Clip 13:14–17:12 (3:58)

Experts create rubrics to evaluate model outputs where humans excel, like lawyers scoring contract redlines. This data supports evals for benchmarking and RL verifiers to reward improvements. The process evolves from supervised fine-tuning to RL from AI feedback, using rubrics for scalable training. Anthropic exemplifies this shift to AI-driven RL.