Garry Tan, CEO of Y Combinator, has expressed support for American AI labs using distillation techniques on frontier AI models, a practice currently under scrutiny due to allegations against Chinese labs. Distillation involves extensively querying a model to understand and replicate its reasoning, a common method for training new AI models. While some, including Anthropic's CEO Dario Amodei, have called for regulatory action against unauthorized distillation, Tan argues that US labs should be free to engage in this practice openly and legally.
Tan emphasizes that proprietary AI labs should not restrict how customers use the information their models provide, comparing this to how these labs themselves trained on vast amounts of public and copyrighted data without explicit permission. He suggests that access to intelligence derived from broadly available data should be treated more like a public good rather than locked behind restrictive terms.
Highlighting the need for a balanced AI ecosystem, Tan supports both frontier AI labs, which push technological boundaries, and open-weight AI labs, which provide broader access and freedom. He warns against the concentration of AI power in a single company, describing it as a potential "doomer scenario" that could stifle innovation and competition.
Tan's stance reflects a call for a regulatory approach that encourages openness and competition in AI development, aiming to prevent monopolization while fostering innovation in the US AI landscape.