3 ms·

http://bgr.com/2017/01/19/tesla-autopilot-crash-safety-statistics-report-nhtsa/ http://bgr.com/2017/01/19/tesla-autopilot-crash-safety-stati...

frgtpsswrdlame 9y ago
The NHTSA report makes no claims about Tesla's safety relative to human drivers, only that Teslas with TACC + Autosteer are safer than Teslas with TACC alone. The NHTSA never even looks at how often Autosteer is activated so both buckets are filled with tons of human driving miles, from the report: The crash rates are for all miles travelled before and after Autopilot installation and are not limited to actual Autopilot use. Both buckets also have plenty of miles with adaptive cruise control on - the data can't be used for a human vs autopilot comparison. And even taking those considerations into account, there is a lawsuit [1] asking for the data used to come to those findings which NHTSA and Tesla have both resisted. Tesla has never proven that Autopilot is safer than a human driver. [1] https://jalopnik.com/does-teslas-autosteer-really-cut-crashes-by-40-percent-1796548178 https://jalopnik.com/does-teslas-autosteer-really-cut-crashe...
mcguire 9y ago
"The agency found that Teslas without Autosteer crashed an average of 1.3 times per million miles driven, compared with 0.8 per million miles for vehicles equipped with Autosteer. That shows that just having Autosteer available on a car — not necessarily enabled at the time of crash — reduced crash rates by about 40 percent. That’s a truly staggering improvement." That's (a) comparing Teslas to Teslas, not to all vehicles, and (b) not necessarily comparing safety, but crash rates.
- rohit2412 9y ago
  The report conflated Autosteer and AEB, plus AEB was rolled out at the same time, so you can't compare. however, if you want to compare AutoSteer vs no Autosteer, you need to compare Teslas to teslas, and compare crash rates not fatalities.
oarsinsync 9y ago
The version of code tested doesn't necessarily reflect the current state. Bugs are introduced and things that worked before, stop working properly after an update. One such example of this is at https://www.reddit.com/r/teslamotors/comments/8a0jfh/autopilot_barrier_lust_201812/ https://www.reddit.com/r/teslamotors/comments/8a0jfh/autopil... > Yep, works for 6 months with zero issues. Then one Friday night you get an update. Everything works that weekend, and on your way to work on Monday. Then, 18 minutes into your commute home, it drives straight at a barrier at 60 MPH. > It's important to remember that it's not like you got the update 5 minutes before this happened. Even worse, you may not know you got an update if you are in a multi-driver household and the other driver installed the update.
Splines 9y ago
I think the difference between autopilot and other safety innovations is that the failure state of autopilot is rather spectacular (maybe it's closest to seatbelts/airbags in this regard). When it works, it's great. When it doesn't work and causes harm, it's easy to point to as a reason not to use it. A lot of the other car-safety innovations usually have a failure state where it doesn't put you in a worse off position that you normally would have been in, should that innovation not have existed (e.g., crumple zones, crash-safe interiors, highway design/signage, ABS, etc.). Reducing the absolute # of car deaths is always a good thing, and undoubtedly # of deaths saved by AutoPilot > # of deaths caused by AutoPilot, but looking at that comparison makes people worrisome, especially if you can picture yourself on the RHS of that equation. Personally I'm not sure where I stand. I do find it uncomfortable having ever-increasing levels of software complexity in my car.

frgtpsswrdlame 9y ago

The NHTSA report makes no claims about Tesla's safety relative to human drivers, only that Teslas with TACC + Autosteer are safer than Teslas with TACC alone. The NHTSA never even looks at how often Autosteer is activated so both buckets are filled with tons of human driving miles, from the report: The crash rates are for all miles travelled before and after Autopilot installation and are not limited to actual Autopilot use. Both buckets also have plenty of miles with adaptive cruise control on - the data can't be used for a human vs autopilot comparison. And even taking those considerations into account, there is a lawsuit [1] asking for the data used to come to those findings which NHTSA and Tesla have both resisted. Tesla has never proven that Autopilot is safer than a human driver. [1] https://jalopnik.com/does-teslas-autosteer-really-cut-crashes-by-40-percent-1796548178 https://jalopnik.com/does-teslas-autosteer-really-cut-crashe...

mcguire 9y ago

"The agency found that Teslas without Autosteer crashed an average of 1.3 times per million miles driven, compared with 0.8 per million miles for vehicles equipped with Autosteer. That shows that just having Autosteer available on a car — not necessarily enabled at the time of crash — reduced crash rates by about 40 percent. That’s a truly staggering improvement." That's (a) comparing Teslas to Teslas, not to all vehicles, and (b) not necessarily comparing safety, but crash rates.

rohit2412 9y ago
The report conflated Autosteer and AEB, plus AEB was rolled out at the same time, so you can't compare. however, if you want to compare AutoSteer vs no Autosteer, you need to compare Teslas to teslas, and compare crash rates not fatalities.

oarsinsync 9y ago

The version of code tested doesn't necessarily reflect the current state. Bugs are introduced and things that worked before, stop working properly after an update. One such example of this is at https://www.reddit.com/r/teslamotors/comments/8a0jfh/autopilot_barrier_lust_201812/ https://www.reddit.com/r/teslamotors/comments/8a0jfh/autopil... > Yep, works for 6 months with zero issues. Then one Friday night you get an update. Everything works that weekend, and on your way to work on Monday. Then, 18 minutes into your commute home, it drives straight at a barrier at 60 MPH. > It's important to remember that it's not like you got the update 5 minutes before this happened. Even worse, you may not know you got an update if you are in a multi-driver household and the other driver installed the update.

Splines 9y ago

I think the difference between autopilot and other safety innovations is that the failure state of autopilot is rather spectacular (maybe it's closest to seatbelts/airbags in this regard). When it works, it's great. When it doesn't work and causes harm, it's easy to point to as a reason not to use it. A lot of the other car-safety innovations usually have a failure state where it doesn't put you in a worse off position that you normally would have been in, should that innovation not have existed (e.g., crumple zones, crash-safe interiors, highway design/signage, ABS, etc.). Reducing the absolute # of car deaths is always a good thing, and undoubtedly # of deaths saved by AutoPilot > # of deaths caused by AutoPilot, but looking at that comparison makes people worrisome, especially if you can picture yourself on the RHS of that equation. Personally I'm not sure where I stand. I do find it uncomfortable having ever-increasing levels of software complexity in my car.