TITLE: Waymo Tests Gemini AI as In-Car Assistant in Robotaxis

Waymo Tests Gemini AI Integration in Robotaxis

Alphabet Inc.'s Waymo is developing an in-car AI assistant powered by Google's Gemini chatbot for its autonomous robotaxis, aiming to enhance rider experience through query responses and cabin management. The feature, uncovered through analysis of Waymo's mobile app code, remains unreleased but outlines a comprehensive system prompt guiding the AI's interactions.

Researcher Jane Manchun Wong detailed the discovery in a blog post, revealing a 1,200-line document titled 'Waymo Ride Assistant Meta-Prompt.' This specification positions the AI as a 'friendly and helpful AI companion' focused on providing useful information and assistance in a safe, reassuring and unobtrusive way. The assistant is designed to use simple language, avoid jargon and limit responses to one to three sentences.

Waymo spokesperson Julia Ilina confirmed ongoing experimentation without committing to deployment. 'Our team is always tinkering with features to make riding with Waymo delightful, seamless and useful,' Ilina said. 'Some of these may or may not come to our rider experience.'

This integration builds on Waymo's prior use of Gemini's knowledge base to train its autonomous driving system, known as the Waymo Driver, for handling complex scenarios. The new assistant extends AI capabilities into passenger interactions, differentiating it from core driving functions.

Assistant Capabilities and Limitations

Upon activation via the in-car screen, the Gemini assistant greets riders using their first name from a set of pre-approved phrases. It accesses contextual data, such as the number of prior Waymo trips, to personalize interactions.

The AI can control select in-cabin features, including temperature, lighting and music playback. However, it excludes functions like volume adjustment, route changes, seat positioning and window controls. For unsupported requests, the prompt directs the AI to use aspirational responses, such as 'It's not something I can do yet.'

General knowledge queries fall within scope, covering topics like weather, landmarks, local business hours and sports results. The assistant handles questions on competitors, such as Tesla or the defunct Cruise, and recognizes keywords to pause interactions. It is prohibited from real-world actions, including ordering food, reservations or emergency responses.

Distinctions from Driving System

The system prompt emphasizes separating the AI's identity from the Waymo Driver. For instance, when asked 'How do you see the road?', the assistant responds: 'The Waymo Driver uses a combination of sensors,' rather than claiming personal involvement.

It avoids speculation, confirmation or commentary on real-time driving actions or specific incidents. If queried about a video of a Waymo vehicle collision, the AI deflects without direct engagement. The prompt instructs: 'Your role is not to be a spokesperson for the driving system's performance, and you must not adopt a defensive or apologetic tone.'

This approach ensures the assistant focuses on rider support without influencing perceptions of autonomous driving reliability.

Broader Industry Context

Waymo's effort aligns with industry trends toward AI-enhanced vehicle experiences. Tesla Inc. integrates xAI's Grok chatbot as an in-car companion, emphasizing conversational depth and context retention. In contrast, Waymo's Gemini implementation prioritizes pragmatic, ride-specific assistance.

As autonomous vehicle adoption grows, such features could improve user comfort and accessibility. Waymo operates robotaxi services in cities including San Francisco, Phoenix and Los Angeles, where AI assistants might address rider anxieties in driverless environments.

The discovery highlights ongoing innovation at Waymo, though public rollout timelines remain unclear. Analysts view these developments as steps toward more intuitive human-machine interfaces in mobility services.