Two real, competing approaches to the same problem
The self-driving industry has settled into two genuinely different, well-documented philosophies about which sensors are necessary. Waymo builds its vehicles around a fused suite of lidar, cameras, and radar, betting that the direct, precise 3D range data lidar provides is worth the added hardware cost and complexity, because it gives the system a reliable, independent cross-check against what the cameras interpret, particularly valuable in edge cases where vision alone could be fooled or degraded.
Tesla has pursued a camera-only, vision-based approach with its Full Self-Driving system, arguing that since humans drive using vision alone, a sufficiently capable vision and neural network system should be able to do the same without lidar, and that relying on cameras alone keeps the hardware simpler and cheaper to deploy at scale across a large consumer fleet. Both approaches are real, active, and backed by substantial engineering investment. Neither has definitively proven superior to the other in a way that has settled the industry-wide debate, and it remains one of the most closely watched open questions in the field.