They can’t automate trains yet, which run in a much more controlled environment than any other transportation method.
What makes them think that they can automate cars, which run in pretty much the most chaotic environment there is? Between erratic drivers, mechanical failures, pedestrians, bicyclists or even squirrels, it’s a wonder the wetware that’s piloting vehicles doesn’t have MORE “incidents” than already occur!
They could automate trains if they really wanted too. Biggest two problems are nobody wants to buy all that new equipment, and them thar unions that require two people in the engine even if neither of them are at the controls.
Train automation is more of a union issue than a technology one.
These cars basically default to “don’t hit anything”, which is pretty easy to do. Combined with “high definition maps” they can even follow a specific lane on a snow covered freeway. They classify most of what they see, including people and some other animals, and can move to a state/speed to anticipate an action, like somebody stepping into the road.
To begin with these cars would probably be limited to specific areas they can handle. There are still certain things they have a hard time with but in general will have a much better safety record than people. Road deaths will drop sharply which is why the NHTSA is playing ball...