3 ms·

I really do not like Tesla's strategy of luring the drivers to doze off while driving. However, the Guardian article on the accident claims that Teslas drove 13

by yk 10y ago

I really do not like Tesla's strategy of luring the drivers to doze off while driving. However, the Guardian article on the accident claims that Teslas drove 130 million miles in autopilot mode, while the rate of fatal accidents is closer to 100 million miles per accident. [0] It is statistically complicated to measure anything from a single event, and ars had a nice article on the problems for self driving cars [1], but I would interpret this as a Tesla autopilot together with the human driver is as good (with some indication of slightly better) than a unassisted human driver. Apart from that, it is important to remember that fatal accident statistics are not just numbers, but are tabulations of individual tragedy. My condolences to his family, but sadly we can not expect autonomous cars to be perfect in the first try. [0] https://www.theguardian.com/technology/2016/jun/30/tesla-autopilot-death-self-driving-car-elon-musk https://www.theguardian.com/technology/2016/jun/30/tesla-aut... [1] http://arstechnica.com/cars/2016/04/car-makers-cant-drive-their-way-to-safety-with-self-driving-cars/ http://arstechnica.com/cars/2016/04/car-makers-cant-drive-th...

frankchn 10y ago
There is only one data point from Tesla so i won't read into it too much either way. That said, all Model S are expensive and fairly new care with good safety features compared to the average car on the road today. Thus,I wonder if it is better to compare the 1 fatality / 130 million miles statistic to Model S (and S-class / 7-series) fatality rates while not on Autopilot mode.
- c0g 10y ago
  A comparison also needs to include the fact that Autopilot is really only used on highway like streets, which are much safer than most other types of driving.
  - mSparks 10y ago
    Not sure i agree. You may be more likely to have an "accident" elsewhere. But im reasonably sure that you are much more likely to die at highway speeds. This is one of the great things about tesla logging everything tho. The data is there to "prove" it one way of the other.
    - c0g 10y ago
      I'm basing my assertions off things like this: http://www.npr.org/2009/11/29/120716625/the-deadliest-roads-are-rural http://www.npr.org/2009/11/29/120716625/the-deadliest-roads-...
      - mSparks 10y ago
        ->The roads traveled least are the nation's deadliest roads Two factors for that tho. Firstly is bad road markings. Which computers will be infinately better at handling. But more importantly is the response time of emergency services and letting someone know theres been an accident. Even minor injuries end up fatal if no one finds you trapped in the car for a week. And every second counts for more serious collisions. Only so much manufacturers can do about either. Dui and driver fatigue. As you will see are among the big killers. (And before you think "yeah but i dont drive drunk or tired" remember you arent the only person on the road). Think its just about being carefull with what is meant by "safer" and "other types of driving". You are more likely to have an accident in busy periods on busy roads. But you are much more likely to die if you have an accident somewhere quiet or on high speed roads.
        c0g 10y ago
        Bad road markings are very hard for current self driving systems to solve. Your other points back up the claim though - self driving cars need to be compared to like-for-like mortality: same roads, same demographic, similar safety features. My guess is they're currently considerably less safe.
- yk 10y ago
  Absolutely, however we can at least conclude with quite a bit of certainty that it is not an order of magnitude worse than human drivers.

by yk 10y ago

frankchn 10y ago

There is only one data point from Tesla so i won't read into it too much either way. That said, all Model S are expensive and fairly new care with good safety features compared to the average car on the road today. Thus,I wonder if it is better to compare the 1 fatality / 130 million miles statistic to Model S (and S-class / 7-series) fatality rates while not on Autopilot mode.

c0g 10y ago
A comparison also needs to include the fact that Autopilot is really only used on highway like streets, which are much safer than most other types of driving.
- mSparks 10y ago
  Not sure i agree. You may be more likely to have an "accident" elsewhere. But im reasonably sure that you are much more likely to die at highway speeds. This is one of the great things about tesla logging everything tho. The data is there to "prove" it one way of the other.
  - c0g 10y ago
    I'm basing my assertions off things like this: http://www.npr.org/2009/11/29/120716625/the-deadliest-roads-are-rural http://www.npr.org/2009/11/29/120716625/the-deadliest-roads-...
    - mSparks 10y ago
      ->The roads traveled least are the nation's deadliest roads Two factors for that tho. Firstly is bad road markings. Which computers will be infinately better at handling. But more importantly is the response time of emergency services and letting someone know theres been an accident. Even minor injuries end up fatal if no one finds you trapped in the car for a week. And every second counts for more serious collisions. Only so much manufacturers can do about either. Dui and driver fatigue. As you will see are among the big killers. (And before you think "yeah but i dont drive drunk or tired" remember you arent the only person on the road). Think its just about being carefull with what is meant by "safer" and "other types of driving". You are more likely to have an accident in busy periods on busy roads. But you are much more likely to die if you have an accident somewhere quiet or on high speed roads.
      - c0g 10y ago
        Bad road markings are very hard for current self driving systems to solve. Your other points back up the claim though - self driving cars need to be compared to like-for-like mortality: same roads, same demographic, similar safety features. My guess is they're currently considerably less safe.
yk 10y ago
Absolutely, however we can at least conclude with quite a bit of certainty that it is not an order of magnitude worse than human drivers.