The brief summary would be that the car detected when an emissions test was being performed, and changed the engine behavior for the duration of the test. Then it changed back to its default, non-compliant, behavior during normal operation.
That’s not news. We knew that already. What’s new about this paper is that it details exactly how it detected emissions tests were happening, and exactly how its behavior changed. Also, they examine how the cheating was done differently in various different makes and models.
I can’t summarize that part because it’s complex and I admittedly don’t fully understand it. But I understand it enough to recognize it is valuable, interesting, and new information.
" This tool is developed for people to get inspired. With playful research AI regenerates street view and aims to show what streets could also look like. Of course AI and the technique, i.e. using Google Streetview images has its limits and is improving moving forward. Feedback and interaction are welcome so we can solve global challenges together. Reach out through cyclinglifestyle@holland.com."
One driverless car ridehail provider in SF shows the safety data from their first million miles and compares it to a human ridehail benchmark. Big chunk of salt since it comes from the company trying to make money from autonomous vehicles. That said, their data is showing that the autonomous vehicles are safer than humans.
Whether this data is real, the company is good, etc. are not topics I care to discuss because I don’t know.
The point is that I think people hold driverless cars to an absurd standard of safety before they are willing to accept them. Until they are perfect, people get irrationally angry about the very idea of them.
That is a wrong way of thinking. IF the autonomous vehicles are safer than human drivers, then we should replace human drivers with them at fast as practically possible.
Cars kill people no matter who or what is behind the wheel. It doesn’t matter to the victim if they get run over by an automatic car or one with a human. What matters is they got run over. If switching to autonomous cars saves lives, even if it doesn’t save all the lives, DO IT.
Wow, that’s a blast from the past, I remember those arguments from years ago, when driverless cars were just as around the corner as they are now.
Also the thing I’d check for is the scale of the numbers, and which numbers they use - The companies selling these things love to use miles driven for safety statistics, because it’s an easily fudged one without technically lying, and they look very impressive. I mean, five million miles, goddamn, that’s a long way, and only 196 crashes!
But, when you treat human miles the same way, it doesn’t work out in their favor, which is why they obfuscate. Five million miles to 196 crashes works out to about 0.004 percent crash rate, rounding up the crash figure a little to 200 to make the math easy. That’s pretty low! Not even one percent!
But human drivers collectively in the US alone drive 8.7 billion miles per day at best estimate, with roughly 20,000 crashes every day(again, rounded up). Of course, to be both fair(since I rounded down to 5 mil) and to make the math easy because I’m a bit lazy, let’s round that down to eight billion, and it works out that human drivers are a literal order of magnitude safer, with a 0.00025 percent crash rate.
Nobody’s asking them to be perfect, at least to begin with, but I’d figure being at least as safe as humans using their own preferred metrics would be a good start.
(And also maybe not causing huge traffic jams in SF and Austin just a short few months apart - the second of which was only days before this report was released.)
To be (more) fair, the robotaxi didn’t cause the accident. A human driver hit a pedestrian, and the robotaxi just responded in a way that happened to be more dangerous than expected.
The human driver who caused the incident hasn’t yet been identified, so the blame ends up on the robotaxi company.
Which I think is actually a good step, but doesn’t go far enough. Imagine if every accident or injury caused by a F-150 means that all F-150’s were banned from the road, or if all drivers had to pay for a second driver to sit next to them, to monitor their driving for safety.
Maybe a remote switch on all vehicles, so at the death of a single pedestrian, all vehicles in the area can’t start the next day?
It’s funny that as soon as an agency has a handy lever to pull in the name of safety, they don’t hesitate to pull it. But if the lever doesn’t exist yet, there’s no comprehension or effort put into wondering if something could be done.
At the very least, I agree it’s a good step. After all, if a machine cannot be held accountable for making life-or-death decisions like a person can, it should not be in a position of making life or death decisions.
At least according to this study, there is a correlation between mobility behavior and positive social attitudes. In other words, people who bike care about the common good and people who drive cars are evil and selfish.
It’s not your imagination. Drivers in Nevada, probably the US, and likely other places around the world, are worse since the pandemic. The sociopathy will add to the enormous death toll beyond the COVID itself.
The cause was easy enough to identify: Data parsed by Kuhls and her colleagues showed that drivers were speeding more, on highways and on surface streets, and plowing through intersections with an alarming frequency. Conversely, seatbelt use was down, resulting in thousands of injuries to unrestrained drivers and passengers. After a decade of steady decline, intoxicated-driving arrests had rebounded to near historic highs.
I routinely have people run red lights long into their timers in front of me in Queens and Manhattan the last few months. Just plow through the intersection without even slowing down at a dead red.
I am reminded of, on another forum, a conversation about self driving cars where someone expressed distaste for them because they would have to maintain a larger follow distance behind AVs because of “AV that is consistently brake-checking those behind them due to nuisance alarms”
I mentioned that minimum safe distances at 65mph is approximately 300 feet, and bet that it was 10 times more what they consider a safe follow distance normally, and still probably five times more than what they were considering when following an AV.
The response that came back was
The problem with the “safe distance” is that it’s more than large enough for a vehicle to merge into, which repeatedly happens.