Autonomous vehicle development is shifting from modular, rule-based systems to "AV 2.0," an approach centered on unified, end-to-end embodied AI. Alex Kendall, co-founder and CEO of Wayve, details how this transition leverages generative AI to move beyond static, map-dependent navigation toward systems capable of generalizing across diverse cities and vehicle types. By training models on massive datasets to predict future states, Wayve enables vehicles to interpret dynamic environments and handle long-tail edge cases without manual intervention. Key innovations include Gaia 1, a generative world model for simulating complex traffic scenarios, and Lingo 1, a driving commentator that utilizes natural language to increase system transparency and user trust. This integration of hardware and software, supported by high-performance compute platforms, aims to create scalable, intelligent autonomous systems that function safely in the real world.
Sign in to continue reading, translating and more.
Continue
