What happened
Waymo, Alphabet's autonomous driving subsidiary, significantly advanced its robotaxi capabilities by integrating AI foundation models into its fifth-generation Driver system, introduced in 2020. This shift from hard-coded rules to AI-driven generalisation enables vehicles to predict pedestrian movements and adapt to new environments, even inferring actions from partially obscured objects. Waymo's system now uses a foundation model, similar to large language models, to power its "driver," a "simulator" for virtual testing, and a "critic" for performance evaluation, with a visual language model trained on Google's Gemini interpreting road conditions.
Why it matters
Robotaxi operators can now scale services more rapidly and cost-effectively due to AI foundation models enabling vehicles to generalise driving experiences across diverse cities and scenarios. This mechanism reduces reliance on extensive, location-specific hard coding and expensive sensor arrays, impacting procurement teams evaluating sensor suites and platform engineers building autonomous systems. The shift allows for faster international expansion and increased operational reach, as evidenced by Waymo's recent scaling of services and testing in new cities like London.
