4 ms·
The "Catch-22" doesn't make a lot of sense, really. In either case you are driving. Until we legally recognise autopilots as being fully in control of the car,
by J-dawg 10y ago
The "Catch-22" doesn't make a lot of sense, really.
In either case you are driving. Until we legally recognise autopilots as being fully in control of the car, the driver is always driving.
Her statement does make me wonder how relaxing it would really be to drive a car with autopilot, though. If I can't sit back and trust it 100%, I think I'd rather just be fully engaged in driving the car.
- gnicholas 10y agoYou're not alone—humans are not very good at constantly monitoring systems, which is essentially what are asked to do in the early days of "autopilot": http://www.csmonitor.com/Business/In-Gear/2016/0827/NASA-thinks-Tesla-Autopilot-is-a-bad-idea http://www.csmonitor.com/Business/In-Gear/2016/0827/NASA-thi...
- J-dawg 10y agoThanks for the link, really interesting. It does seem that there's a huge amount to overcome as autopilot moves from driver aid to fully autonomous. I'm starting to think the ideal compromise (for now) is a car where I'm in control, but with lots of systems that cut in when I make a mistake, rather than the other way around.
- vkou 10y agoIf you don't take control from autopilot when it does something dangerous, Tesla blames you for not paying attention. When you do take control, Tesla blames you, because you disengaged autopilot. Either way, autopilot, by definition, didn't have an accident. This lets them claim that it has driven XYZ miles without an accident - by ignoring the 150 feet that preceded one (As the human took control - even if not taking control would have resulted in the same accident.)
- M_Grey 10y agoRead back what you just wrote, and then consider the impact that might have on any statistics which seek to describe the safety of the Autopilot. By definition you've just decided that it's always driver error, which would skew those statistics.
- J-dawg 10y agoWhat I'm trying to say is that it's not appropriate to classify accidents as either caused by autopilot or human error. The fatal accident was, tragically, a failure of both the autopilot and the driver. But the driver was relying on the autopilot in a way that has never been recommended. It does seem as though Tesla has been overstating autopilot's safety, based on arguments made in the article and in the comments here. But it seems grossly unfair to judge the autopilot against a use case that it was never recommended for. It's a bit like braking at the last second on a wet road and then blaming your ABS when you crash. I would love to see some more nuanced analysis of autopilot safety when compared to an appropriate control group.
- M_Grey 10y agoI think that nuanced analysis is important, and maybe until it exists companies such as Tesla should be more careful regarding safety claims.