Edtech Insiders AI Doesn’t Learn Alone: Dr. Magdalena H. Gross of Mercor on Human Intelligence
Mar 30, 2026
Dr. Magdalena H. Gross, a Stanford-trained AI education leader who helps experts translate deep knowledge into AI-ready systems and runs community arts work. She explains how experts train AI through structured thinking. She discusses making invisible thinking visible, parallels between human learning and reinforcement learning, spiral curricula for scaling skill, and why interdisciplinary thinking matters.
AI Snips
Chapters
Transcript
Episode notes
Human Experts Power Large Language Models
- Human experts are the backbone of modern LLMs, providing domain-specific knowledge that models lack from raw web data.
- Magdalena Gross trains domain experts (e.g., fruit fly geneticists) to externalize how they think so models can learn their procedures and judgments.
Fruit Fly Geneticist Learned To Teach A Model
- A fruit fly geneticist couldn't explain his tacit knowledge until Gross asked step questions about working with other insects.
- He became an expert tasker and later a research scientist after learning to externalize his process.
AI Training Mirrors Human Reinforcement Learning
- Training has moved from simple preference labeling to forms of reinforcement learning that mirror human learning.
- Experts give iterative feedback on hundreds or thousands of model responses, reinforcing good outputs and correcting errors so models generalize better.