I don't think Tesla has any sort of talent edge in the self-driving car space. Their original autopilot was mostly built on tech from MobileEye and from what I've heard Tesla doesn't pay great either, and unlike SpaceX their mission is not unique.
Having a network of cars collecting data is great, but they may not actually be able to collect all of it simply because it is super high bandwidth, whereas those with test cars can stick a rack of disks in the car and collect all of the data.
Does Tesla not pay well by industry standards, or is that when comparing them to software companies? Seems like they work their employees very hard, though, so maybe they lose out on talent– on the other hand there was a time when Apple worked its employees hard and they were still able to retain top people.
Regarding the data collection, that's a great point. I hope that's not the case though, as I would think the data is far more valuable than the bandwidth it's carried on. Though, I suppose the capacity simple doesn't exist to send it over the air the way you can just store locally and physically offload. Then I guess it comes down to whether the compressed data from an enormous number of vehicles still gives a significant advantage.
Having a network of cars collecting data is great, but they may not actually be able to collect all of it simply because it is super high bandwidth, whereas those with test cars can stick a rack of disks in the car and collect all of the data.