The System Behind Self-Driving: Waymo’s Dmitri Dolgov

In this episode of The a16z Show, Waymo co-CEO Dmitri Dolgov shares the technical and strategic evolution behind Waymo’s fully autonomous driving system. From its origins in Google’s self-driving project in 2009 to now operating over 500,000 fully autonomous rides weekly across 11 U.S. cities, Waymo has transitioned from a research-focused moonshot to a global scaling operation. Dolgov explains the core architecture of the Waymo Driver, which relies on a multi-sensor fusion of LiDAR, radar, and cameras, processed through a foundation model that powers specialized off-board systems: the driver, simulator, and critic. These systems enable safe, scalable, and socially aware driving by combining end-to-end AI with structured intermediate representations. He emphasizes that full autonomy cannot be achieved through incremental improvements to driver-assist systems, and highlights the breakthroughs in AI, simulation, and hardware—especially the sixth-generation Ojai platform—that have made large-scale deployment possible. Dolgov also discusses the future of mobility, including the potential for autonomous ride-hailing to transform urban landscapes by reducing parking demand and improving traffic flow. Key takeaways include: 1) The Waymo system is built on a foundation model that powers specialized AI teachers (driver, simulator, critic), distilled into efficient on-board models. 2) Sensor fusion—LiDAR, radar, and cameras—provides complementary, robust perception, with LiDAR excelling in high resolution and radar in adverse weather. 3) True full autonomy requires a fundamentally different architecture than driver-assist, not an incremental upgrade. 4) The sixth-generation Ojai vehicle and sensor stack are simpler, cheaper, and more scalable, enabling deployment across diverse cities and vehicle platforms. 5) The future of mobility lies not just in self-driving cars but in reimagining urban design, with autonomous fleets reducing the need for parking and improving traffic efficiency. Dolgov expresses excitement about global expansion and the transformative potential of AI-driven world models in simplifying and accelerating autonomous systems.