A second option for Waymo

Waymo got a problem, Airzona wants a driver. I can help out.

Arizona should instead ask for watchers,  human agents nearby who can stop the car in trouble.  This lets Waymo go after profitable venues where they can work in clustors, sufficient to cover the cost of the watcher.

Second,  prequal the passengers.  Use a membership system where frequent users git a bit trained on the emergency stop button, they take some driver risk.  Some frequent users can take the whole risk, be acceptable drivers on their own.  This is good in local, quiet neighborhoods where residents take trips locally.

Finally, get the cars talking. As more cars have the cameras and features detectors, make it a regulatory law that they communicate obstacles and circumstances to each other.  Define a standard traffic swarming protocol.  Usually the taxibot will be around one or two other cars watching the road, they should be required to communicate, legally.  There is no bias here, the communication anonymous. In a situation, one bot may call a stop on another.

Legally, a car manufacture cannot have cameras that detect emergencies and fail to warn other bots.  There is actually a human law about reporting emergencies and extends via power of attorney to the bots. So, simply having the federal government mandate swarming for safety is enough, Waymo and the rest are obligated to make the standard.

IOmagine the mercedes bot discovering a dangerous car up the road, it pulls over flashing red. The dangerous car spins around the corner and whallops ford bot.  The mercedes passenger says, yea, my car warned me.  But didn't warn ford bot just behind the curve? It is an uproar, a bad marketing position for the industry.  These cars are proliferating, they need connecting.

These three things increase the watcher technology so increasing the bots over time.