Y Combinator CEO Tan says he would not take action against AI model distillation
Y Combinator startup accelerator CEO Garry Tan said during the annual Demo Day that he would not take action against the distillation of artificial intelligence models. CNBC reports.
Distillation is the use of responses from a more powerful AI model to train a smaller or less powerful model; in some cases, this practice may be illegal. It has become a concern for developers of frontier models, including OpenAI and Anthropic. Anthropic accused Chinese companies Moonshot AI, DeepSeek and MiniMax of distillation. OpenAI, for its part, believes that the architectures of DeepSeek V3 and R1 were distilled from its GPT-4 and GPT-4o models.
Balance between models
In Tan’s view, regulators should focus less on curbing distillation and instead seek a balance between open-weight models and frontier models. He believes frontier models should retain a pricing advantage that would allow their business model to remain viable, while open models would provide users with freedom and access.
Tan acknowledged that achieving such a balance is difficult, calling it “walking a tightrope.” At the same time, he is convinced that this approach could deliver the best outcome.
More current news is available on the UA.News Telegram channel Telegram.
Urgent AI risks
The Y Combinator chief also called for focusing on the current risks of artificial intelligence rather than scenarios involving humanity’s extinction. He stressed the need to rely on “scientific facts, not science fiction” and called cybersecurity an immediate threat.
On Thursday, Anthropic said it had blocked access to Claude in five cases in which scientists from unnamed foreign countries used the models to research dangerous pathogens. The company feared that the researchers might have covertly used Claude to create biological weapons.
Tan also believes that job losses caused by AI and economic changes will not have an immediate effect and will unfold over decades. At the same time, Y Combinator is actively investing in this sector: of the 196 startups that presented at Demo Day, 149 were classified as machine learning and artificial intelligence projects.
