3 ms·

Yeah how many have they saved to date? It’s a valid point and I’d love to know the answer now that we know self-driving vehicles are responsible for >= 1 person

Yeah how many have they saved to date? It’s a valid point and I’d love to know the answer now that we know self-driving vehicles are responsible for >= 1 person’s death

anonym29 3y ago
I don't know how you'd even measure the numbers of deaths that would've resulted from accidents that didn't happen because an autonomous car avoided it in a situation where an inattentive, reckless, or unskilled human driver wouldn't have... especially without an otherwise identical control environment. Or even situations where partial self-driving (such as Tesla's system) reduced driver fatigue that would've later contributed to a more exhausted driver causing an accident. I'm the furthest thing from a Tesla fanboy, but I did try one for a while in order to have a more unbiased view and I will say that "autopilot" subjectively reduced my stress and fatigue during and after 5 mph bumper to bumper traffic jams and left me feeling more attentive afterwards compared to those same types of traffic conditions in cars being operated 100% by me.
- rgmerk 3y ago
  One of the problems with trying to do this with fatalities is that fatalities are actually pretty rare per passenger/mile. Consequently, the amount of testing required to prove that the fatality rate of autonomous vehicles is lower than human-driven one’s is astronomical. Non-fatal accidents are more common but the statistics are far more rubbery.
  - taeric 3y ago
    I'd assume that a large part of the data that makes up the "mile" portion of the passenger per mile is large trips. So, look at traffic deaths in some dense cities and you will probably find much larger numbers than you'd expect. As a quick check of this idea, looking at "vehicle passenger deaths atlanta" in a search gives numbers around 100 for many years. (And Atlanta isn't that dense, all told.)
- ummonk 3y ago
  Being able to handle 5 mph bumper to bumper traffic jams is hardly a Tesla-specific feature. It's quickly becoming a standard feature in cars, and such driver assistance features are rather independent from the conversation about full self-driving (whether via Cruise cars or Tesla FSD).
- taeric 3y ago
  In normal circumstances, this would be easy. Just look at the base rate of deaths and see how the needle moved after these were introduced. This is how we know, for example, that automatic emergency brakes are saving lives. Unfortunately, the pandemic through normal out the window. :( Probably still possible to find some natural experiments where you can control some variables. My gut would be that their are some comparable cities that differ mainly in whether they have these cars. That said, it will be a lot harder, as the impact of the pandemic was pretty massive.

Yeah how many have they saved to date? It’s a valid point and I’d love to know the answer now that we know self-driving vehicles are responsible for >= 1 person’s death

anonym29 3y ago

I don't know how you'd even measure the numbers of deaths that would've resulted from accidents that didn't happen because an autonomous car avoided it in a situation where an inattentive, reckless, or unskilled human driver wouldn't have... especially without an otherwise identical control environment. Or even situations where partial self-driving (such as Tesla's system) reduced driver fatigue that would've later contributed to a more exhausted driver causing an accident. I'm the furthest thing from a Tesla fanboy, but I did try one for a while in order to have a more unbiased view and I will say that "autopilot" subjectively reduced my stress and fatigue during and after 5 mph bumper to bumper traffic jams and left me feeling more attentive afterwards compared to those same types of traffic conditions in cars being operated 100% by me.

rgmerk 3y ago
One of the problems with trying to do this with fatalities is that fatalities are actually pretty rare per passenger/mile. Consequently, the amount of testing required to prove that the fatality rate of autonomous vehicles is lower than human-driven one’s is astronomical. Non-fatal accidents are more common but the statistics are far more rubbery.
- taeric 3y ago
  I'd assume that a large part of the data that makes up the "mile" portion of the passenger per mile is large trips. So, look at traffic deaths in some dense cities and you will probably find much larger numbers than you'd expect. As a quick check of this idea, looking at "vehicle passenger deaths atlanta" in a search gives numbers around 100 for many years. (And Atlanta isn't that dense, all told.)
ummonk 3y ago
Being able to handle 5 mph bumper to bumper traffic jams is hardly a Tesla-specific feature. It's quickly becoming a standard feature in cars, and such driver assistance features are rather independent from the conversation about full self-driving (whether via Cruise cars or Tesla FSD).
taeric 3y ago
In normal circumstances, this would be easy. Just look at the base rate of deaths and see how the needle moved after these were introduced. This is how we know, for example, that automatic emergency brakes are saving lives. Unfortunately, the pandemic through normal out the window. :( Probably still possible to find some natural experiments where you can control some variables. My gut would be that their are some comparable cities that differ mainly in whether they have these cars. That said, it will be a lot harder, as the impact of the pandemic was pretty massive.