Interview, Fireside Chat
How End-to-End Learning Created Autonomous Driving 2.0: Wayve CEO Alex Kendall
- Aims to become a universal embodied AI foundation model capable of generalizing across applications and diverse sensor architectures, including camera-only, radar, and lidar configurations, to amortize costs and enable rapid adaptation.
- Plans to deploy across hundreds of cities globally within months of hardware identification, leveraging transfer learning to minimize data requirements and utilize next-generation onboard compute like the NVIDIA Thor chip for large language and vision-language models.
- Targets a transition from driver assistance to "eyes-off" autonomy and driverless robotaxis, seeking to eliminate the 95% of accidents currently caused by human error while achieving superhuman performance levels through deep OEM integration.
- Forecasts a market shift where mass-produced vehicles from top manufacturers include GPUs, surround cameras, radar, and front-facing lidar, creating a significant opportunity to partner with the 90 million cars built annually.
- Anticipates the rollout of tens to hundreds of thousands of affordable robotaxis by avoiding hardware retrofitting in favor of native software integration across passenger vehicles, trucking, and broader robotics verticals.
- Predicts the evolution of "AV 3.0" where intelligence moves outside the vehicle to enable V2X communication, potentially eliminating traffic lights, and may restrict human driving to designated recreational areas due to the inability to communicate with autonomous mesh networks.
- Plans to expand into manufacturing, manipulation, and humanoid robotics by scaling general-purpose low-cost hardware stacks, driven by compounding returns in data, compute, algorithms, and embodiment.
- Identifies engineering efficiency, superior measurement systems for regression detection, and high-fidelity simulators that close the real-world gap as critical competitive advantages.
- Notes that while safety and flow generalize well, utility factors such as navigation, road semantics, language, and driving cultures remain the primary challenge for global expansion.
- Expects the software-defined infrastructure market to grow from automotive into other robotics sectors, emphasizing the need for scalable computing to support open platforms for AI.