If AI is suitable for 99% of driving, why not use remote drivers for the remaining 1%? (eg. accidents, construction, adverse weather) You could have a central

If AI is suitable for 99% of driving, why not use remote drivers for the remaining 1%? (eg. accidents, construction, adverse weather) You could have a central office in a cheap location of the USA for professional drivers to take control as required. We already do it military for drone strikes...