However, Waymo is Deep Blue of self-driving cars. Doing very well in a closed space. As a result of this geofencing, they have effectively exhausted their search space, hence they work well as a consequence of lack of surprises.
AI works well when search space is limited, but General AI in any category needs to handle a vastly larger search space, and they fall flat.
At the end of the day, AI is informed search. They get inputs, and generate a suitable output as deemed by their trainers.
> hence they work well as a consequence of lack of surprises. Emphasis mine.
In this context, "lack of surprises" is exactly the rest of the driving besides road choice. In the same space, the behaviors of other actors are also a finite set, or more precisely, can be predicted with much better accuracy.
I drive the same route for ~20 years for commute. The events which surprise me are few and far between, because other people's behavior in that environment is a finite set, and they all behave very predictably, incl. pedestrians, bikes, and other drivers.
Choosing roads are easy, handling surprises hard, but if you saw most potential surprises, then you can drive even without thinking. While I'm not proud of it, my brain took over and drove me home a couple of times on that route when I was too tired to think.