"Fatalities per distance driven" is a terrible metric unless FSD is doing all the driving for the trips the data is pulled from. Otherwise what tends to happen

"Fatalities per distance driven" is a terrible metric unless FSD is doing all the driving for the trips the data is pulled from. Otherwise what tends to happen is FSD/autopilot drives the low-risk highway miles and the human has to take over in the high-risk city miles, skewing the stats.