Y Combinator’s Garry Tan wants U.S. open-weight AI labs to ‘distill’ frontier models, too
AI-summarised brief · reviewed before publication
Garry Tan, CEO of Y Combinator, told CNBC and TechCrunch that U.S. AI labs should be free to employ “distillation” on American frontier models, mirroring techniques used by Chinese labs. Distillation involves prompting a larger model extensively to extract its reasoning patterns for training smaller, open‑weight systems. Tan argues that restricting how customers use API outputs oversteps proprietary labs’ rights, especially since those labs have already trained on vast public data without permission. He rejects calls for regulatory crackdowns, proposing instead a voluntary American distillation regime to broaden open‑weight alternatives and prevent a monopoly by a single proprietary provider. Tan emphasizes that open‑weight models foster freedom, access, and a healthier competitive landscape in AI development.
💡 Why It Matters
- · By championing open‑weight distillation, Tan aims to diversify AI power sources and curb the risk of a single dominant, closed‑source AI monopoly.
