Inception CEO Stefano Ermon: diffusion LMs generate text 10x faster than autoregressive models
No Priors · youtube · 2026-10-01
On the No Priors podcast, Inception Labs CEO Stefano Ermon discusses the 2024 breakthrough paper proving diffusion models could match traditional autoregressive LLMs on quality while generating text 10x faster — a result strong enough to start the company. The interview covers the diffusion-LM technical approach and the commercialization rationale.
More from Companies & People
- Creative Machines returns to NYC on Oct 15 with fireside chat and live demo — round · 2026-10-01
- WSJ: Jensen Huang and Tech CEOs Privately Pushed Back on Amodei's AI Risk Warnings — robleclerc · 2026-10-01
- Weekly visual AI/ML interview map: from ML fundamentals to RAG take-homes — datawithsuman · 2026-10-01
- Was Google's announcement of an unusable product just perf-review season theater? — aronchick · 2026-10-01
- Clean Code author says he no longer reads code: the goal was always to free up thinking — julsimon · 2026-10-01
- SeatGeek's company-wide ChatGPT rollout spawned 300 IT tickets on day one — aakashgupta · 2026-10-01