Waymo executive argues Tesla's camera-only system can’t achieve full autonomy

Waymo’s senior software leader, Srikanth Thirumalil, published a blog titled “10 AI Lessons from Driving 200+ Million Fully Autonomous Miles,” in which he argued that a camera-only perception stack cannot support safe, fully autonomous operation at scale. Citing Waymo’s extensive mileage, he emphasized that combining cameras with lidar and radar creates a redundant, rich world model that a single sensor cannot match. Thirumalil contrasted this with Tesla’s strategy, which dismisses lidar as unnecessary and relies heavily on cameras and high-detail maps, a stance he says limits the system’s robustness.

He also described Waymo’s use of continuously updated HD maps as a supplemental “mental memory” for complex maneuvers. The piece notes that New Jersey’s pending robotaxi legislation would mandate multiple sensors, effectively barring Tesla’s camera-only vehicles. Finally, Thirumalil warned that upgrading a Level 2 driver-assist system to full autonomy is a false goal, insisting that true Level 4 capability requires purpose-built hardware and extensive real-world testing.