Dmitri Dolgov: Waymo and the Future of Self-Driving Cars | Lex Fridman Podcast #147

In this episode, Lex Fridman sits down with Dmitri Dolgov, the CTO of Waymo, to explore the current state and future of autonomous vehicle technology. The conversation begins with Dolgov's personal journey, including his childhood interests in computer games and robotics that eventually led him to pursue advanced technical work at Moscow Institute of Physics and Technology. He recounts his early involvement in the DARPA Urban Challenge, a pivotal competition that helped establish many of the foundational techniques now used in autonomous driving.

Dolgov provides detailed insights into Waymo's origin story as Google's self-driving car project and explains the sophisticated hardware architecture underlying modern autonomous vehicles. He discusses the various sensors including lidar, cameras, and radar systems that work together to perceive the environment. A significant portion of the episode focuses on Waymo's operational service in Phoenix, where the company has deployed a fully driverless ride-hailing service without human safety drivers. Dolgov explains how this real-world operation generates invaluable feedback that informs continuous improvements to the system.

The discussion touches on connected cars and infrastructure considerations, exploring how vehicles might communicate with each other and their environment. Dolgov emphasizes the importance of creating products that people genuinely love, noting that successful autonomous vehicles must be reliable, safe, and provide a compelling user experience. He addresses the philosophical question of whether self-driving cars need to break traffic rules like human drivers sometimes do, arguing for strict adherence to safety protocols while adapting intelligently to real-world conditions.

Waymo's expansion into autonomous trucks represents another frontier in the technology, with significant implications for the logistics industry. Dolgov explains how the principles of autonomous driving scale to larger vehicles and the unique challenges they present. The episode delves into the critical role of lidar technology in autonomous driving, comparing it with camera-only approaches and discussing the advantages and limitations of different sensor modalities.

A crucial theme throughout the conversation is the necessity of machine learning for handling the infinite variability of real-world driving scenarios. Rather than relying solely on pre-programmed rules, modern autonomous systems must learn from vast amounts of data to make safe decisions in novel situations. Dolgov discusses pedestrian safety and addresses the famous trolley problem, explaining that while ethical dilemmas are philosophically interesting, the practical focus is on preventing accidents altogether rather than optimizing outcomes in unavoidable crashes.

The episode concludes with personal reflections on the meaning of life and Dolgov's book recommendations, offering a broader perspective on technological progress and human purpose. Throughout the conversation, Dolgov demonstrates deep technical knowledge while remaining accessible to a general audience, providing a comprehensive view of where autonomous vehicle technology stands and where it is heading.