3 ms·
It is the classic trolley[1] problem and Tesla is flipping the switch on the tracks. I don't think there is anything to indicate that Tesla's approach is leadi
by slg 9y ago
It is the classic trolley[1] problem and Tesla is flipping the switch on the tracks. I don't think there is anything to indicate that Tesla's approach is leading to more deaths. Considering there has only been two confirmed fatalities in which autopilot was partially at fault (AFAIK), I would bet that autopilot is still well in the positives in "lives saved" versus "lives ended". Whether that is the right ethical or moral decision is something that can be debated forever, but Musk and in turn Tesla have made it very clear where they stands on it.
[1] - https://en.wikipedia.org/wiki/Trolley_problem https://en.wikipedia.org/wiki/Trolley_problem
- txcwpalpha 9y agoBut this isn't a question about no autopilot vs autopilot. This is a question about the previous version of autopilot vs the current version. If the previous version of autopilot (read: the one that didn't swerve into barriers and kill someone) was working, why did they update everyone to this new version which apparently has a life-threatening bug? Was it necessary to roll out this update so quickly? My guess is no, and that with less of a "move fast and break things" mentality, they might have tested the update more and fixed this issue before it went live.
- slg 9y agoThe same principle applies. Autopilot is not a binary feature that either works or doesn't. It is an evolving system that should continuously get better and safer. Sure, Tesla could have delayed this specific update in favor of more testing. But how many preventable accidents would happen during that testing period that an earlier roll out could have prevented? Neither you or I know the answer to that. Tesla has access to all the autopilot data and therefore should be able to more accurately predict that number. I haven't seen anything to suggest Tesla is ignoring that data or being reckless in when they decide to roll out these changes.
- txcwpalpha 9y ago>It is an evolving system that should continuously get better and safer. In this instance, it didn't get safer. It did the opposite, and it killed someone. >I haven't seen anything to suggest Tesla is ignoring that data or being reckless in when they decide to roll out these changes. You haven't? Because I'm sitting here looking at a story about a bug in Tesla's latest software that Tesla ignored and then it resulted in someone's death.
- slg 9y agoYou are drawing those conclusions with only half of the equation. Like I said in my previous comment, neither of us know how many accidents were prevented or lives saved by that update.
- txcwpalpha 9y ago>Like I said in my previous comment, neither of us know how many accidents were prevented or lives saved by that update. Then why are you making assumptions about it, then? >You are drawing those conclusions with only half of the equation. No, I'm not. The facts are: there was a problem that was reported to Tesla, Tesla ignored that problem, and now someone is dead because of that problem. No further information is needed to recognize that Tesla acted poorly here.
- slg 9y agoI said Tesla failed in my first comment. I am not defending them regarding this particular accident. I am simply explaining the potential logic behind their decisions and why this single failure doesn't mean their approach is inherently wrong. The only assumption I am making is that Tesla has good intentions with their aggressive autopilot releases.
- fmihaila 9y ago> and now someone is dead because of that problem. You've stated this repeatedly without proof. It has not been established he is dead because of that problem. The investigation may yet reach that conclusion, but until then you have no support for this assertion.
- throwaway2048 9y agothere is also no indication that it is _saving_ lives either, stats that are about easy, relatively safe situations, where humans take over when challenges occur are not any kind of meaningful indicator. Not to mention the market telsa courts self selects for safer driving in general (rich, and informed). To act like this is a clearcut case of the trolley dilemma and telsa is making a moral choice that saves life is misinformed bordering on disingenous. There is a huge amount of indication that Telsa is acting both irresponsibly (refusing to admit fault, when its clear fault exists, refusing to address concerns, only answering with deflections), and recklessly (misleading marketing about fully capable autodriving hardware, calling a system that requires constant, careful manual attention "autopilot", enabling it by default, absolutely refusing to back down even a little)
- slg 9y agoAs far as I know there has been only a single public study on the safety of Tesla's autopilot. It compared miles driven in Teslas without the autopilot hardware and safety features versus miles driven in Teslas with those features. They showed a 40% reduction in crashes. I think it is fair to assume that a 40% reduction in crashes resulted in numerous lives being saved. I don't see how coming to that conclusion is "misinformed bordering on disingenous". That specific note is available on page 10-11 in the linked study. [1] [1] - https://www.scribd.com/embeds/337007075/content?start_page=1&view_mode=scroll&access_key=key-DgnugphNmXkiFVNAWAvV https://www.scribd.com/embeds/337007075/content?start_page=1...
- rohit2412 9y agoThis isn't limited to no autopilot vs full autopilot. If autopilot is so good at correcting a drunk driver's accident, then let it run in a shadow mode only making corrective actions in emergencies. Why is AEB bundled in with Autosteer in all stats? AEB is undoubtedly good, Autosteer is debatable.
- taneq 9y ago> AEB is undoubtedly good, Autosteer is debatable. AEB is clearly unreliable, as it didn't automatically slow the car in any of these cases where the car was pointed directly at a significant-sized stationary solid object. I did worry about this sort of thing back when they mentioned nuisance trips of the radar-based AEB system, and that to deal with them they were basically just geotagging where these nuisance trips happened and disabling the system near those points. I'd like to know if AEB was actually active at the location where the fatality occurred.
- rohit2412 9y agoAEB is there to correct driver's mistakes, not drive the car itself. Yes, AEB fails, but a driver + AEB is safer than a driver alone.
- unityByFreedom 9y ago> I don't think there is anything to indicate that Tesla's approach is leading to more deaths. The safety claims Tesla/Musk make are debatable. There is a lawsuit over getting the raw data NHTSA used to make the "autosteer reduces collisions by 40%" [1]. AEB also reduces the collision rate by 40 percent, which is mentioned in the same NHTSA report [2]. And, supposedly AEB and autosteer were activated around the same time, so it might not be easy to tell which provided a benefit. > I would bet that autopilot is still well in the positives in "lives saved" versus "lives ended" You need a lot of data to show that. 1 death per 86 million miles driven is the average in the US, and that's across all types of vehicles, new and old, including motorcycles, bad weather (where autopilot is not likely to be engaged), and on all kinds of roads (autopilot is mostly engaged on divided highways). There was also a death in China that looks like autopilot [3]. [1] http://www.safetyresearch.net/blog/articles/quality-control-systems-corp-sues-dot-tesla-data http://www.safetyresearch.net/blog/articles/quality-control-... [2] https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF [3] https://www.youtube.com/watch?v=fc0yYJ8-Dyo https://www.youtube.com/watch?v=fc0yYJ8-Dyo