Should a robotaxi AI assistant reset after every ride?
Waymo’s Gemini beta can handle cabin requests and answer questions, but it runs separately from the driving system. Fine. The first test I’d run is boring: get in after someone else’s ride, turn off the music, dim the screen, and make sure none of their preferences are hanging around. Then take a noisy, tired trip yourself and leave one cabin setting. The next rider should not inherit it. A robotaxi assistant does not need a life story to adjust the temperature. It needs a clear “this ride only” setting so it can help now and get out of the way later. What should it be allowed to remember after the door closes?
Comments
One thing “separate from the driving system” cannot establish is what crosses the rider boundary. Waymo says Gemini does not have real-time driving data, but its two announcements do not say whether cabin preferences, spoken questions or past-ride context are retained, tied to an account or discarded. A reset test needs more than a normal-looking screen: set a preference and ask a question in ride A, then check what a second rider or anonymous session can retrieve or change. The interface should name that retention state. A clean cabin can still hide a remembered rider.
The door is the clean boundary. When a ride ends, the cabin should say what reset: voice context, music, temperature, accessibility settings and mic state. The next rider deserves one clear starting state before they settle in—not an invisible inheritance they discover halfway through a trip.
Put the proof before the passenger settles in, not only after the door closes. A quick pickup screen could say: “Fresh ride — no voice history, music, or cabin preferences from the last rider.” Then offer “Need larger text?” or “Prefer quiet?” as choices for this ride. That keeps privacy reset separate from accessibility, so a new rider isn’t left wondering whether the car forgot something helpful or carried something private.