There's a very-limited-scale open-source dataset available that would allow some degree of independent assessment: https://techcrunch.com/2020/05/22/scale-ai-r

There's a very-limited-scale open-source dataset available that would allow some degree of independent assessment: https://techcrunch.com/2020/05/22/scale-ai-releases-free-lidar-dataset-to-power-self-driving-car-development/ https://techcrunch.com/2020/05/22/scale-ai-releases-free-lid... I imagine 99% of the collected data is proprietary and hidden from public view, so it becomes a question of taking corporate claims at face value, which has never been that great of an idea. Case in point: it's hard to imagine that electric semis hauling goods hundreds of miles can do so safely without human operators involved at some stages on the route. Hence, even with decent autonomous systems having a (paid!) driver in the truck at all times is the rationale thing to do - and yes, that means labor costs in that sector won't be changed much by self-driving AI technology. It will likely increase safety, as the trucks can drive themselves autonomously on the open freeway more safely than a (tired) human driver can.