[ad_1]
Waymo appears to be testing adding Google’s Gemini AI chatbot to its robotaxis in an effort to integrate an AI assistant that would guide riders and answer their questions, according to findings from researcher Jane Manchun Wong.
“While digging through the code of Waymo’s mobile app, I discovered the full system prompt for the unreleased Gemini integration,” Wong wrote in a blog post. “The document, internally titled ‘Waymo Ride Assistant Meta-Prompt’, is a more than 1,200-line specification that defines exactly how the AI assistant is expected to behave in a Waymo vehicle.”
The feature isn’t yet available in public builds, but Wong says the system prompt makes it clear that this is « more than a simple chatbot. » The assistant could answer questions, manage certain cabin functions such as climate control and, if necessary, reassure passengers.
“While we don’t have details to share today, our team is always tinkering with features to make driving with Waymo delightful, seamless and useful,” Julia Ilina, a Waymo spokesperson, told TechCrunch. “Some of these may not reflect our rider experience.”
This wouldn’t be the first time Gemini has been integrated into Alphabet’s self-driving business stack. Waymo says it has used Gemini’s « world knowledge » to train its autonomous vehicles to navigate complex, rare and demanding scenarios.
Wong writes that the assistant is tasked with having a clear identity and purpose: “a friendly and helpful AI companion integrated into an autonomous Waymo vehicle” whose primary purpose is “to improve the rider’s experience by providing useful information and assistance in a safe, reassuring and unobtrusive manner.” The bot is instructed to use clear, simple language and avoid technical jargon, and is instructed to keep its responses concise in one to three sentences.
According to the system prompts, when a rider activates the assistant via the in-car screen, Gemini can choose from a series of pre-approved greetings personalized with the rider’s first name. The system also has access to contextual data about the rider, such as how many Waymo rides they have taken.
Techcrunch event
San Francisco
|
October 13-15, 2026
The prompts currently allow Gemini to access in-car functions such as temperature, lights and music. Notably absent from the features list are volume control, route changes, seat adjustment and window controls, Wong points out. If a rider asks for a feature that Gemini has no control over, the bot should respond with « aspirational phrases, » such as, « It’s not something I can do yet. »
Interestingly, the assistant is instructed to make a clear distinction between its identity as Gemini, the AI bot, and the autonomous driving technology (the Waymo Driver). So when you answer a question like, “How do you see the road?” Gemini should not say, « I’m using a combination of sensors, » but should instead respond, « The Waymo Driver uses a combination of sensors… »
The system prompts include a series of fascinating tidbits, such as how the bot should handle questions about competitors like Tesla or the now-defunct Cruise, or which keywords will cause it to stop talking.
The assistant should also avoid speculating about, explaining, confirming, denying or commenting on real-time driving actions or specific driving events. So if a passenger asks about a video they’ve seen of a Waymo hitting something, the bot is instructed not to answer right away and to deflect.
“Your role is not to be a spokesperson for propulsion system performance, and you should not adopt a defensive or apologetic tone,” the prompt reads.
The assistant in the car may answer general knowledge questions, such as about the weather, the height of the Eiffel Tower, what time the local Trader Joe’s closes and who won the last World Series. You are not allowed to take real-world actions such as ordering food, making reservations, or handling emergencies.
Waymo isn’t the only company integrating AI assistants into self-driving vehicles. Tesla is doing something similar with xAI’s Grok. However, the two different car assistants have different functions. Gemini seems programmed to be more pragmatic and ride-oriented, while Grok is presented more as a buddy in the car who can have long conversations and remember the context of previous questions.
[ad_2]
Source link
Comments