Overview
- Waymo co‑CEO Dmitri Dolgov told a Y Combinator audience that camera-only systems can match human driving in many situations but stop improving long before they reach the reliability needed to operate without a human.
- Waymo runs separate processing for cameras, lidar and radar and then fuses those outputs so the system gets color and detail from cameras, 3D geometry from lidar, and direct velocity readings from radar.
- Dolgov cited Waymo test examples where lidar detected a pedestrian in a Phoenix dust storm when camera feeds were nearly unusable, showing how non‑camera sensors work in darkness, glare, dust, fog, rain and snow.
- Company-reported mileage highlights the operational gap used to argue progress: Waymo says it has logged about 220 million rider-only miles while Tesla’s public tally of driverless miles is roughly 380,000.
- Dolgov framed safety as an “exponential ladder of nines,” meaning each extra increment of reliability needs roughly ten times more work, a reality that shapes how quickly regulators and cities may accept full driverless service.