Garry Tan, CEO of Y Combinator, has expressed support for American AI labs employing distillation techniques similar to those used by Chinese AI developers. Distillation involves prompting advanced AI models extensively to understand and replicate their reasoning, a common practice in AI training. Tan argues that U.S. labs should be free to use these methods openly, without regulatory interference.
This stance contrasts with recent concerns raised by Anthropic, which accused Chinese labs of conducting unauthorized distillation attacks using fraudulent credentials. Anthropic's CEO, Dario Amodei, has called for regulatory action against such practices. However, Tan emphasizes that American labs should engage in distillation transparently and with permission.
Tan's position is grounded in the belief that proprietary AI labs should not restrict how users interact with their models. He points out that many AI models were initially trained on vast amounts of publicly available data, often without explicit consent from content owners. Therefore, restricting downstream use of model outputs could be seen as overly limiting.
He envisions a balanced AI ecosystem where frontier labs continue to innovate and receive funding, while open-weight models provide broader access and freedom to users. Tan warns against a scenario where a single dominant company controls the most advanced AI technologies, which could stifle competition and innovation.
By advocating for open distillation practices, Tan aims to foster a more diverse and accessible AI landscape in the U.S., countering the dominance of any single proprietary entity.