Waymo联合CEO德米特里·多尔戈夫在Y Combinator创业学校的演讲中指出，仅依靠摄像头的自动驾驶系统无法实现完全自主驾驶。[1]多尔戈夫表示，虽然人类可以仅用眼睛驾驶，但弱传感会导致安全曲线过早平坦化，无法达到自主驾驶所需的可靠性标准。[1]相比之下，Waymo采用的三传感器融合方案整合摄像头、激光雷达和雷达，在220万英里无人驾驶测试中表现显著优于人类驾驶员。[1]
根据Waymo的最新安全报告，该公司的系统导致的严重伤害事故比人类驾驶员少94%，性能优势约为17倍。[1]而特斯拉过去一年的无人驾驶累计里程为380,000英里，该公司在2021年撤除了雷达，在2022年撤除了超声波传感器。[1]多尔戈夫强调，达到自主驾驶所需的可靠性需要经过指数级的提升。[1]仅有单个传感器的覆盖缺陷就可能导致系统完全瘫痪——比如一片树叶覆盖传感器即可让机器人停止运行。[1]目前Waymo每周为全美15个城市提供约50万次付费行程服务，其完整解决方案（包括车辆、传感器套件和驱动程序）成本约为60,000美元。[1]
Waymo co-CEO Dmitri Dolgov outlined the fundamental limitations of camera-only autonomous driving systems in a presentation at Y Combinator's Startup School, arguing that reliance on a single sensor modality inherently constrains safety performance.[1] Dolgov noted that while humans can drive using eyes alone—"there's that proof of existence"—this does not translate to reliable machine autonomy, as "weak sensing just leads to a safety curve that flattens out way too early."[1] By contrast, Waymo's multi-sensor approach combining cameras, lidar, and radar has logged 2.2 million driverless miles and demonstrated a safety record 94 percent better than human drivers in serious injury accidents, roughly 17 times superior to human performance.[1]
The distinction between the two approaches reflects fundamentally different engineering philosophies.[1] Tesla has progressively stripped away sensor redundancy, removing radar in 2021 and ultrasonic sensors in 2022, accumulating 380,000 autonomous miles over the past year.[1] Dolgov highlighted the practical fragility of camera-only systems: a single leaf covering a sensor can cause a robot to halt entirely.[1] Achieving the reliability required for autonomous vehicles demands what he termed "an exponential ladder of nines" in safety certification.[1] Waymo currently provides approximately 500,000 paid robotaxi trips weekly across 15 U.S. cities, operating a complete system—vehicle, sensor suite, and software stack—estimated at roughly $60,000 per unit.[1]