Joe Spisak interviews Ion Stoica at the AI and Data at Scale event, discussing reinforcement learning, large language model evaluations, and the evolution of open-source projects. Stoica shares his journey from networking to AI, emphasizing the importance of community building and collaboration between academia and industry. They delve into the challenges of managing open-source projects like VLLM, the significance of de facto standards in AI, and the shift towards distributed training. The conversation also touches on the LAMP stack of this generation, the explosion of agentic frameworks, and the need for low-level APIs in rapidly evolving fields.
Sign in to continue reading, translating and more.
Continue