Y Combinator's Garry Tan says 'do nothing' about distillation (www.cnbc.com)

Garry Tan, CEO of Y Combinator, recently expressed a nonchalant stance regarding the distillation of AI models by Chinese competitors, asserting that he would "do nothing" about the issue. During Y Combinator's Demo Day, Tan emphasized the importance of focusing on immediate risks associated with AI, such as cybersecurity threats, rather than sensational existential fears. He noted that although distillation—where less capable models are trained on the outputs of more advanced ones—raises concerns for frontier model developers like OpenAI and Anthropic, he believes the regulatory focus should be on achieving a balance between open prototype models and frontier models, allowing for fair competition without stifling innovation. Tan's perspective is significant as it diverges from the prevailing narratives within Silicon Valley and national security circles, where calls for regulatory action against such distillation practices are intensifying. He highlighted the complex legal landscape surrounding data usage in AI training, referencing ongoing lawsuits over copyright infringement, while advocating for a pragmatic approach to AI safety that prioritizes current dangers over speculative risks. This balance, he argues, is crucial as the AI landscape continues to evolve, with Y Combinator actively investing in AI startups—149 out of the 196 presented at Demo Day—indicating ongoing confidence in the sector’s growth potential despite the challenges.