If someone sitting in the vehicle can lose attention, there is not much hope for someone watching from afar on a TV (even if you're streaming at something like 4k 120fps --- which would be an immense amount of bandwidth to handle for each car --- it's still not much more immersive than watching someone playing a driving game.)
Maybe they can make it more like a video game, where you're constantly saving cars from collisions, and you don't know which is a simulation and which is real. Give them bonuses based on what percentage of collisions they successfully deter.
And let's have Facebook moderators be shown a small percentage of images known to be of rape or beheadings just to test if they're flagging images correctly. You see how that might be a bad idea?