4 ms·

That's a general problem with most news about accidents or deaths. They cherry pick the interesting stories, not the significant ones. The way I see news is tha

That's a general problem with most news about accidents or deaths. They cherry pick the interesting stories, not the significant ones. The way I see news is that it's like fiction but with the extra excitement of being real, much like watching a movie that's "based on a true story". Its purpose is entertainment, not information or understanding of the world. Nobody should be making a decision to buy a Tesla or not based on stories like these. I recently mentioned self-driving cars to somebody and her response was "Oh, one killed a woman recently, didn't it?" That was about the extent of what she knew about them because she watches the news.

mikeash 8y ago
Yes, I say that any risk that makes the news is probably not worth worrying about. What I don’t get is all the HN commenters acting like a few crashes proves that Autopilot is irredeemably dangerous.
- DannyBee 8y ago
  Because it's way oversold for what it does, and even if you look at other traffic aware cruise control, all data suggests it is in fact, more dangerous than other traffic aware cruise control (probably in part because of the former) I can't find an NTSB investigation or study that pegs traffic aware cruise control as the cause of the accident in any case except in Tesla's cases. Of course, it seems to Tesla folks this is just evidence they are gunning for Tesla instead of evidence that maybe it is actually worse :)
  - mikeash 8y ago
    What data is there? I didn’t realize there was enough to reach a conclusion yet.
    - DannyBee 8y ago
      Start with: https://www.ntsb.gov/news/press-releases/Pages/PR20170912.aspx https://www.ntsb.gov/news/press-releases/Pages/PR20170912.as... Also, 18 crashes so far where autopilot was enabled? vs so far, literally zero known crashes across everyone else? Do you not find it somewhat interesting that in zero other cases has the NTSB found that TACC or the design of a TACC played a role in the crash? I'd also point out: the supplier of their original crash tech broke up with them and ordered them to stop using it because they felt what was happening was unsafe (see Mobileye's letter to Tesla). Tesla later tried to spin it as "they are made we are breaking up with them" but Mobileye then produced data showing tried to renew the contract and Mobileye refused. In fact, plenty of other car companies are building self-driving tech, and currently relying on Mobileye, and Mobileye has not refused to supply them - only Tesla. Like, i'm not sure how this set of stuff doesn't tilt a person into "hey, maybe this is something to be worried about, they should produce data why it isn't".
      - mikeash 8y ago
        I'm extremely skeptical that there have been zero TACC-involved crashes with other cars. Has the NTSB or other organization actually put out a statement to that effect somewhere, or are you just basing this on an inability to locate any when searching?
        DannyBee 8y ago
        Again, please provide any data that says otherwise. At all. Any! So far the data in this conversation has been incredibly one sided, with me providing data, and you providing skepticism. These systems exist in a lot of cars, yet nothing until Tesla suggests any issues anywhere (i'm ignoring things like news reports which are fairly biased) Therefore, the extraordinary claim is not that Tesla is worse, it's that Tesla isn't worse.
        cherioo 8y ago
        A quick search gives me this http://www.thetruthaboutcars.com/2016/08/automatic-emergency-braking-wont-always-stop-crash-americans-think-will/ http://www.thetruthaboutcars.com/2016/08/automatic-emergency.... In particular "At 45 mph, the crash-avoidance systems slowed the test vehicle by 74 percent, on average, and prevented collisions 40 percent of the time.", i.e. it fails 60 percent of the time. Do you really believe other cars' automatic emergency braking works 100 percent of the time with perfect reliability? They don't show up because it's non-news.
        DannyBee 8y ago
        No, i don't believe that, i just would like this person to provide literally any data at all for their pet theories instead of just crapping on everyone else's imperfect data (which is pretty typically the way of the world) So yeah, what i want is a citation or anything other than smug hand waving and dismissivenessm, to actually make it a useful discussion.
        mikeash 8y ago
        The "data" you've provided so far consists of: 1) A link to IIHS crash avoidance test results. The Model S got a perfect score in those results, so those results don't at all support any claim that the Model S is worse. (I can certainly accept that the test is not definitive and the Model S could be worse than others while still getting a perfect score. My point is merely that this is not relevant data.) 2. A mention of 18 crashes with Autopilot enabled. Without a baseline to compare it to or at least some sort of rate based on the total fleet activity, this is more of an anecdote than data. Note that the average American driver crashes about once every 18 years (https://www.forbes.com/sites/moneybuilder/2011/07/27/how-many-times-will-you-crash-your-car/#2f35cda24e62 https://www.forbes.com/sites/moneybuilder/2011/07/27/how-man...) and there are around 175,000 Teslas on American roads right now (https://www.greencarreports.com/news/1115454_tesla-expects-200000th-u-s-sale-this-year-electric-car-tax-credits-to-start-phaseout https://www.greencarreports.com/news/1115454_tesla-expects-2...). If Tesla drivers crash at the average rate, we'd expect to see about 175000/18/365 = 26.6 Tesla crashes every day, and that's just in the US. 18 Autopilot crashes ever is... not very high. 3. A completely unsourced and frankly implausible claim that the crash rate for other manufacturers' driver assistance features is zero.

mikeash 8y ago

Yes, I say that any risk that makes the news is probably not worth worrying about. What I don’t get is all the HN commenters acting like a few crashes proves that Autopilot is irredeemably dangerous.

DannyBee 8y ago
Because it's way oversold for what it does, and even if you look at other traffic aware cruise control, all data suggests it is in fact, more dangerous than other traffic aware cruise control (probably in part because of the former) I can't find an NTSB investigation or study that pegs traffic aware cruise control as the cause of the accident in any case except in Tesla's cases. Of course, it seems to Tesla folks this is just evidence they are gunning for Tesla instead of evidence that maybe it is actually worse :)
- mikeash 8y ago
  What data is there? I didn’t realize there was enough to reach a conclusion yet.
  - DannyBee 8y ago
    Start with: https://www.ntsb.gov/news/press-releases/Pages/PR20170912.aspx https://www.ntsb.gov/news/press-releases/Pages/PR20170912.as... Also, 18 crashes so far where autopilot was enabled? vs so far, literally zero known crashes across everyone else? Do you not find it somewhat interesting that in zero other cases has the NTSB found that TACC or the design of a TACC played a role in the crash? I'd also point out: the supplier of their original crash tech broke up with them and ordered them to stop using it because they felt what was happening was unsafe (see Mobileye's letter to Tesla). Tesla later tried to spin it as "they are made we are breaking up with them" but Mobileye then produced data showing tried to renew the contract and Mobileye refused. In fact, plenty of other car companies are building self-driving tech, and currently relying on Mobileye, and Mobileye has not refused to supply them - only Tesla. Like, i'm not sure how this set of stuff doesn't tilt a person into "hey, maybe this is something to be worried about, they should produce data why it isn't".
    - mikeash 8y ago
      I'm extremely skeptical that there have been zero TACC-involved crashes with other cars. Has the NTSB or other organization actually put out a statement to that effect somewhere, or are you just basing this on an inability to locate any when searching?
      - DannyBee 8y ago
        Again, please provide any data that says otherwise. At all. Any! So far the data in this conversation has been incredibly one sided, with me providing data, and you providing skepticism. These systems exist in a lot of cars, yet nothing until Tesla suggests any issues anywhere (i'm ignoring things like news reports which are fairly biased) Therefore, the extraordinary claim is not that Tesla is worse, it's that Tesla isn't worse.
        cherioo 8y ago
        A quick search gives me this http://www.thetruthaboutcars.com/2016/08/automatic-emergency-braking-wont-always-stop-crash-americans-think-will/ http://www.thetruthaboutcars.com/2016/08/automatic-emergency.... In particular "At 45 mph, the crash-avoidance systems slowed the test vehicle by 74 percent, on average, and prevented collisions 40 percent of the time.", i.e. it fails 60 percent of the time. Do you really believe other cars' automatic emergency braking works 100 percent of the time with perfect reliability? They don't show up because it's non-news.
        DannyBee 8y ago
        No, i don't believe that, i just would like this person to provide literally any data at all for their pet theories instead of just crapping on everyone else's imperfect data (which is pretty typically the way of the world) So yeah, what i want is a citation or anything other than smug hand waving and dismissivenessm, to actually make it a useful discussion.
        mikeash 8y ago
        The "data" you've provided so far consists of: 1) A link to IIHS crash avoidance test results. The Model S got a perfect score in those results, so those results don't at all support any claim that the Model S is worse. (I can certainly accept that the test is not definitive and the Model S could be worse than others while still getting a perfect score. My point is merely that this is not relevant data.) 2. A mention of 18 crashes with Autopilot enabled. Without a baseline to compare it to or at least some sort of rate based on the total fleet activity, this is more of an anecdote than data. Note that the average American driver crashes about once every 18 years (https://www.forbes.com/sites/moneybuilder/2011/07/27/how-many-times-will-you-crash-your-car/#2f35cda24e62 https://www.forbes.com/sites/moneybuilder/2011/07/27/how-man...) and there are around 175,000 Teslas on American roads right now (https://www.greencarreports.com/news/1115454_tesla-expects-200000th-u-s-sale-this-year-electric-car-tax-credits-to-start-phaseout https://www.greencarreports.com/news/1115454_tesla-expects-2...). If Tesla drivers crash at the average rate, we'd expect to see about 175000/18/365 = 26.6 Tesla crashes every day, and that's just in the US. 18 Autopilot crashes ever is... not very high. 3. A completely unsourced and frankly implausible claim that the crash rate for other manufacturers' driver assistance features is zero.