Ep. 17: Stefano Ermon, Stanford CS Professor and CEO of Inception Labs

The podcast explores Stefano Ermon's journey in AI, from his electrical engineering background to his current work on generative models. Ermon discusses his early research in automatic theorem proving and his shift towards data-driven methods, including generative adversarial networks and diffusion models. He highlights the development of diffusion models as a more principled approach to image generation, which has influenced technologies like Midjourney and SORA. Ermon also details his work at Inception Labs, focusing on diffusion language models and their applications in autocomplete systems and code generation. The conversation touches on the challenges and rewards of balancing academic research with industry applications, and the qualities that define a successful researcher.