Oils4AsphaultOnly wrote:GRA wrote:Since you love studies so much, here's an NHTSA report claiming a 40% reduction in airbag deployments with Autopilot v1 (the report is embedded in the article): https://electrek.co/2017/01/19/tesla-cr ... lot-nhtsa/
And it's directly about Tesla, not some side-effect study that you keep pulling up.
Terrific. Now let's have Tesla release all their data which Elon has claimed show that A/P-operating Teslas are safer than non-A/P cars. Professional statisticians pointed out the numerous methodological flaws behind his claims at the time he made that statement, and Consumer Reports and other auto safety organizations have asked for that data to be released. I'll be perfectly happy to acknowledge that semi-autonomous systems such as A/P have lead to an overall reduction in accidents (if not a reduction in accidents that A/P is responsible for) if the data is validated by an independent entity and shown to be scientifically valid.Oils4AsphaultOnly wrote:You've once asked why AEB didn't stop for the parked fire trucks, page9 of the report cites a BMW explanation showing why EVERY AEB system is diliberately defeated when the vehicle exceeds a certain speed. They don't want the cars automatically braking from full speed for false-positives. The more people understand this, the fewer crashes into parked firetrucks and center dividers there would be.
As I noted previously (maybe in other topics) every TACC/AEB system is currently unable to handle this sort of event reliably because they are unable to recognize real positives among the false ones, and as such fails to meet the necessary safety requirements. Since people will continue to use the systems improperly either due to misunderstanding their capabilities (which leads to automation complacency), such systems are simply too ineffective to be safe for use by the general public. As has been previously mentioned, lack of understanding of system capability has been and is a major problem in automation-involved aviation accidents, even among highly-trained commercial/military pilots, never mind the much less qualified and trained general driving public. Until the level of idiot-proofing for these systems is much higher than it currently is, they don't belong in the public sphere. As an extreme example of automation complacency: