Is this another selective comparisons article? Autopilot is primarily a highway tool, but the vast vast majority of accidents happen on normal streets, and is

Is this another selective comparisons article? Autopilot is primarily a highway tool, but the vast vast majority of accidents happen on normal streets, and is generally unavailable in many of the cases where accidents are common. A comparison of crash rates is only relevant when comparing similarly situated crashes. So what is the average American crash rate on highways in clear weather, without construction, etc? I suspect that crash rate is far better than at night, during a blizzard, on an ordinary poorly lit suburban road. Yet AP wouldn't run then.