I expect "no human drivers permitted on public roads" to be the law because human drivers are fallible. Robots are also fallible, but they tend to accumulate their lessons across entire robot population (at least within one manufacturing company) whereas each new human driver needs to be taught anew. Even well-taught, humans are still prone to errors and it's only a matter of time before standard robot is 10 times safer than 95th percentile of drivers.
The only problem with robots is systemic risk - some sort of common bug (e.g. Y2K) that would take out many machines at once or another thing that fails in a non-graceful manner. Car companies are not strangers to designing proper failure modes so I think it will pan out ok in the end, but software engineers aren't as good at it yet so we do have that risk for a while.
The only problem with robots is systemic risk - some sort of common bug (e.g. Y2K) that would take out many machines at once or another thing that fails in a non-graceful manner. Car companies are not strangers to designing proper failure modes so I think it will pan out ok in the end, but software engineers aren't as good at it yet so we do have that risk for a while.