Waymo Taps Gemini After 200 Million Miles of Autonomous Driving

Waymo is using more than 200 million miles of autonomous-driving experience to codify 10 lessons for developing AI at scale. Its software is evolving from a collection of specialized models toward a larger foundation-model architecture, while retaining cameras, lidar and radar rather than relying on vision alone. The Alphabet unit also keeps independent safety-verification systems in place, reflecting the higher standard required when model errors can create physical risks on public roads.

In its latest disclosure, Waymo said its autonomous system had surpassed 200 million miles and described a split between fast, low-latency driving models and Gemini-powered reasoning for rare, ambiguous or multi-step situations. The company characterizes the latter as “slow thinking,” allowing larger models to analyze complex scenarios without taking over every real-time task. The approach puts generative AI closer to the driving stack, but Waymo said multimodal sensing and separately validated safeguards remain essential.