The Amazon Go situation was also wildly misrepresented in media, to be clear. It's fairly obvious that they did actually have some vaguely accurate video processing tech, it's just that the reliability never hit a level that the cost of fixing up errors actually saved money vs the alternative.

(The same consideration also applies to Waymo: even if they are not controlling the car like a RC car, does the cost of running their interventions turn the unit economics of their business upside-down? And if not, would this still be true if they were paying US wages for it?)