Waymo’s top software executive has publicly challenged the camera-only approach to autonomous driving, a position widely associated with Tesla and its CEO, Elon Musk. In a blog post published Wednesday, Srikanth Thirumalai, Waymo’s VP of Onboard Software, wrote that cameras are not sufficient for safe, fully autonomous operations at scale. He argued that after more than 200 million real-world miles, the data shows that a redundant sensor suite is required. Waymo combines cameras with lidar and radar to build what Thirumalai described as a rich, redundant world view that no single sensor can replicate.
The post does not mention Musk or Tesla by name, but it directly counters the argument that cameras alone can achieve full autonomy. Musk has long dismissed lidar as a crutch and has predicted that companies relying on it are doomed, reasoning that humans drive with their eyes and so should autonomous vehicles. Tesla has recently doubled down on this stance, with its head of AI, Ashok Elluswamy, arguing on a recent earnings call that safe and affordable autonomy is achievable with cameras alone, dismissing the need for lidar, radar, and HD maps. This puts Tesla largely alone among major autonomous vehicle operators, as Waymo, Zoox, and Motional all use lidar and radar alongside cameras.
The sensor debate has a concrete policy angle in the United States. New Jersey’s state legislature is considering a bill that would legalize robotaxis but only if they include multiple sensors, a requirement that would bar Tesla from operating under such a law. Thirumalai also addressed mapping, another area where Waymo and Tesla diverge. Musk has said high-detail maps are a bad idea, believing autonomous vehicles should rely on instant analysis. Waymo, however, treats HD maps as a powerful prior that helps jump-start validation and acts as a mental memory for the vehicle, using an AI-driven system to keep them continuously updated.
Thirumalai used the post to outline additional principles, including the value of closed-loop simulation for identifying edge cases and the usefulness of Vision-Language Models as reasoning tools. He concluded with his most pointed swipe at Tesla, stating that there is no substitute for actual fully autonomous miles and that improving a Level 2 driver-assist system is a false summit. He argued that true Level 4 maturity can only be achieved through a purpose-built system, validated on closed courses and hardened by driving without a human in the car.
This strikes at the core of Tesla’s strategy, which holds that its Autopilot and Full Self-Driving (Supervised) systems will eventually evolve into fully autonomous platforms through enough data and training. Waymo currently operates fully autonomous vehicles in 11 cities, conducting 500,000 paid trips weekly. Tesla has its own robotaxis, some supervised by human safety drivers and an increasing number unsupervised, and is preparing to launch its purpose-built Cybercab. Musk remains clearly trailing Waymo despite years of claims about his company’s autonomous capabilities, and whether he will adjust his approach is uncertain.
More AI news from TechManNews.
