During Y Combinator's annual Demo Day, Garry Tan expressed a nonchalant stance towards the ongoing debate over AI model distillation, a process where outputs from advanced AI models are used to train less capable ones.
While companies like OpenAI and Anthropic have raised alarms about Chinese firms allegedly engaging in this practice, Tan believes that regulators should prioritize creating a balance between open weight models and frontier models, rather than imposing restrictions on distillation.
He highlighted the ongoing legal disputes surrounding AI training data, including lawsuits involving The New York Times and Anthropic, which underscore the complexities of copyright in AI. Tan also emphasized the importance of addressing immediate cybersecurity risks over speculative fears about AI's future, advocating for a focus on current threats rather than hypothetical scenarios.
He noted that while AI may lead to job displacement, the transition towards automation will unfold over decades, allowing society to adapt. Y Combinator remains committed to investing in AI, with a significant majority of startups at their Demo Day focused on machine learning and AI technologies