Supervised Learning and Reinforcement in AI

Al Engineering 101 with Chip Huyen (Nvidia, Stanford, Netflix)

Topic 3 of 7

Published Oct 23, 2025 · Clip 15:23–18:21 (2:59)

The conversation covers supervised learning using labeled data for tasks like spam detection, with companies like Mercor and Scale providing high-quality labels for labs. Chip discusses reinforcement learning's role in post-training, using human feedback (RHF) to train reward models that guide better outputs. Verifiable rewards, like math problems with exact answers, and AI feedback are also used; economics favor labs with multiple data providers.