Waymo Tests Google's Gemini AI as Companion for Robotaxi Riders

Credits: Waymo

Alphabet's Waymo unit has begun testing Google's Gemini artificial intelligence model as an assistant inside its self-driving vehicles, according to code discovered in the company's mobile app.

The integration, uncovered by researcher Jane Manchun Wong, involves a detailed system prompt that outlines how the AI would interact with passengers during rides.

The assistant aims to handle cabin controls and answer questions while keeping a strict separation from the vehicle's driving functions.

The code reveals a 1,200-line document titled "Waymo Ride Assistant Meta-Prompt" that specifies the AI's behavior.

This prompt directs Gemini to manage features such as temperature, lighting and music playback inside the robotaxi.

It also allows the AI to respond to general inquiries, such as the height of the Eiffel Tower or the winner of the World Series, but prohibits actions like ordering food or altering the vehicle's route.

For unsupported requests, the system instructs the AI to reply with phrases like "It’s not something I can do yet."

One excerpt from the prompt states:

"You are Gemini, a friendly and helpful AI companion integrated into a Waymo autonomous vehicle."

The AI must distinguish itself from the autonomous driving system, known as the Waymo Driver, by redirecting queries about navigation or sensors to that technology.

For instance, if asked how it sees the road, the correct response is:

"The Waymo Driver uses a combination of sensors like lidar, cameras, and radar to see the world around it."

Waymo operates its robotaxi service in cities including San Francisco, Phoenix and Los Angeles, where riders summon vehicles through the app for autonomous trips.

The Gemini feature remains unreleased in public versions of the app, and the company has not announced a rollout date.