3 ms·
It's worth pointing out that the frame of the argument: that this is horrifyingly dangerous behavior, is actually kinda belied by the facts on the ground. The
by ajross 6y ago
It's worth pointing out that the frame of the argument: that this is horrifyingly dangerous behavior, is actually kinda belied by the facts on the ground. The car wasn't used properly, but it did in fact drive itself safely.
People need to be aware of this kind of circular logic, which was applied in the Texas crash debates too: the fact that there are still safety controls in place on a new feature that isn't quite ready for full autonomy doesn't constitute evidence that the feature is unsafe. That require an entirely different kind of evidence.
And frankly... while everyone keeps circling around the same minutiae about the driver nag... Teslas on autopilot aren't actually crashing. In fact it's been quite a long time now since we've seen a juicy autopilot failure.
- aaomidi 6y agoAs someone who owns a Model 3, I call bullshit on 40000 miles without being in the driver seat. Seriously no.
- ghaff 6y agoIt really seems implausible. So he sat in the back seat with a foot on the steering wheel for more than 50 24-hour days of driving or the equivalent of about 6 round trips from cost to coast in the US? And we pretty much know that Autopilot won't work under all conditions.
- maxerickson 6y agoDo we know if there were any near misses? That's how I evaluate my own driving, I think of things much less severe than collisions as incidents.
- ajross 6y agoDoes e.g. the NTSB track "near misses" for human drivers? That would no doubt be great data to have, but it's impractical to collect for all known technologies, even ones widely regarded as having acceptable safety. Thankfully in this case cars actually crash a lot, so we have good proxy data anyway. And... again, Tesla seems to be doing pretty well. People are desperate for bad news about autopilot, to the extent that we all blew up a story about (what turns out to have been) a routine human driver crash in Texas two weeks ago based on a quote from one fire marshall. But the reality is that the system appears to be doing pretty well.
- maxerickson 6y agoMy point is that "The car wasn't used properly, but it did in fact drive itself safely." isn't really something we can know. We know it didn't collide with anything.
- ajross 6y agoRight, and my point was that this is a specious distinction. It's not a knowable thing for any vehicle, anywhere. So the lack of data for one particular vehicle says basically nothing of any use. Like it or not, we measure safety in vehicle travel by measuring accidents. Those are objectively identifiable and there are extensive bureaucracies both public and private designed to track them. And quite frankly, Teslas on autopilot look pretty damn good.
- velcrovan 6y agoYeah, it’s been since April, when a Tesla drove itself into a tree and exploded, killing everyone inside. https://www.wired.com/story/fatal-crash-renews-concerns-teslas-autopilot/ https://www.wired.com/story/fatal-crash-renews-concerns-tesl...
- ajross 6y agoSigh. A lie, around the world, yada yada. Per the NTSB report, autopilot couldn't be engaged on that road, and the driver was seen on his home security video entering the driver seat just 550 feet from the crash site. As everyone knowledgeable was screaming at the time, that just wasn't an autopilot accident. It got spun that way because one firefighter on scene decided (incorrectly) that "no one was driving the car".