Garry Tan, CEO of Y Combinator, has expressed support for American AI labs using distillation methods to learn from leading AI models, similar to practices reportedly used by Chinese labs. Distillation involves extensively querying a model to understand its reasoning and behavior, a common technique for training new AI models. While some, like Anthropic CEO Dario Amodei, have called for regulatory action against what they describe as unauthorized distillation by Chinese entities, Tan argues against imposing restrictions on such practices in the US.
Tan clarified that he does not endorse illicit activities such as using stolen credentials for distillation but believes that legitimate, open access should be encouraged. He views attempts by proprietary AI labs to limit how users interact with their models as overly restrictive. Tan also points out that these labs initially trained their models on vast amounts of publicly available data, often without explicit permission from content owners.
He advocates for a balanced AI ecosystem where frontier labs continue to innovate and receive funding, while open-weight models provide broader access and freedom for users. Tan warns against a scenario where AI development becomes monopolized by a single company with dominant resources and researchers, which he considers detrimental to the industry.
By encouraging open distillation, Tan envisions a more competitive and diverse AI landscape in the US, reducing reliance on foreign models and fostering innovation across multiple players.