Waymo rejects Tesla's cameras-only approach, pushes for higher safety standards
Waymo has rejected Tesla's approach of using only cameras for self-driving cars, arguing that safety standards must exceed human driving capabilities. Srikanth Thirumalai, Waymo's VP of onboard software, emphasized the need for more sensors, while Tesla's AI chief, Ashok Elluswamy, believes cameras alone are sufficient. Waymo currently uses 29 cameras, 5 lidars, and 6 radars in its robotaxis, with plans to reduce to 13 cameras and 4 lidars by 2026. Despite the cost implications, Thirumalai insists on maintaining high safety standards, acknowledging the ongoing challenge of defining what constitutes "safe enough."
Waymo isn’t buying Tesla’s pitch that self-driving cars should work like human drivers. Tesla says cameras alone should be enough, the same way people use their eyes to drive.
Srikanth Thirumalai, Waymo’s vice president of onboard software, disagrees. He told Business Insider the standard needs to be higher than human driving.
Ashok Elluswamy, Tesla’s AI chief, spoke at the ScaledML Conference on January 29 and framed self-driving as an AI problem, not a sensor problem. His argument: humans navigate with eyes, so autonomous cars should manage with cameras.
Thirumalai runs over 600 engineers building Waymo’s software. He’s not interested in that approach.
Nobody knows what ‘safe enough’ actually means
The hardware tells you everything. Tesla wants fewer than 10 cameras and AI trained on billions of driving miles. Waymo robotaxis carry 29 cameras, five lidars, and six radars. About 2,500 Waymo vehicles operate across U.S. cities now. The next version coming by late 2026 drops to 13 cameras, four lidars, and six radars. Still keeping lidar.
The tension is cost versus safety. More sensors cost more money, which makes it harder to scale to millions of vehicles. Fewer sensors might create safety problems that regulators and riders won’t accept. Thirumalai said Waymo decides what safety level it needs, then figures out how to cut sensor costs and improve the software. He thinks the setup will change in three to five years, but won’t drop lidar just because it’s expensive.
What counts as safe enough? Nobody really knows. Thirumalai admitted Waymo is still working that out. They don’t promise robots will be twice or five times safer than humans. They look at specific driving situations, check how often they happen per million miles, then try to beat that rate.
Waymo co-CEO Tekedra Mawakana already said a robotaxi will eventually kill someone. It’s not if, it’s when.
Safety data shows stark differences in crash rates
Waymo told senators its cars had 10 times fewer serious crashes than human drivers over the same distance, per Cryptopolitan’s earlier reporting. That data came from an independent audit covering 200 million autonomous miles. Tesla reported its Full Self-Driving cars average 5.1 million miles between major crashes. The national average for human drivers is 699,000 miles.
Morgan Stanley analyst Adam Jonas said Waymo’s tech works but costs more than Tesla’s camera system. The price difference matters when both companies need tens of thousands of cars to match millions of human Uber drivers.
Videos keep showing up online of autonomous vehicles screwing up in school zones, around emergency vehicles, in bad weather, during regular drives. Thirumalai said expecting AI to never make mistakes isn’t realistic.
Want your project in front of crypto’s top minds? Feature it in our next industry report, where data meets impact.
