How End-to-End Learning Created Autonomous Driving 2.0: Wayve CEO Alex Kendall

In this interview, Alex Kendall, CEO of Wayve, discusses the evolution of autonomous driving (AV) from AV 1.0, which relied on hand-engineered code and high-definition maps, to AV 2.0, which utilizes end-to-end neural networks and general-purpose AI. Wayve's approach focuses on building intelligent robots capable of generalizing and adapting to various vehicles, sensor architectures, and use cases without heavy infrastructure. The conversation covers the challenges and architectural considerations of ensuring safety and real-time performance, the importance of data diversity and world models for reasoning and generalization, and the integration of large language models for improved representations and human-robot interaction. Kendall also touches on Wayve's go-to-market strategy of partnering with automotive OEMs, the shift from driver assistance to eyes-off autonomy, and the potential for AV 3.0 to involve vehicle-to-vehicle communication and shared intelligence.