Y Combinator's Garry Tan Calls for US Labs to Distill Open-Weight AI Models

Garry Tan, president and CEO of startup accelerator Y Combinator, has called on American open-weight artificial intelligence labs to adopt model distillation techniques used to train systems based on frontier AI models, according to a report from TechCrunch.

Distillation is a training method in which a smaller AI model is taught using outputs from a larger, more capable model, allowing developers to produce more efficient systems without the extensive computing resources required to train frontier models from scratch.

According to TechCrunch, Tan wants smaller American open-weight AI labs to apply this approach using U.S.-based frontier AI labs as the source material. The goal, per the report, is to give the United States a broader and more competitive set of open-weight AI options that are not developed in China.

Open-weight models, which make their underlying parameters publicly available for developers to build upon, have become a point of competition between U.S. and Chinese AI developers. Chinese labs have released a number of open-weight models that have gained attention for their performance and accessibility, according to industry observers.

TechCrunch's report frames Tan's comments as part of a broader conversation within the U.S. tech industry about how domestic AI companies can maintain competitiveness in the open-weight model space relative to international rivals.

Further details on Tan's specific proposals, timeline, or which frontier labs might be involved were not included in the available report.