Yes, it's pretty well understood at this point that news reports often greatly overstate the impact of an event compared to base rates. By playing to the "local neighborhood" and "cute kitty" angle while pointing at the "obvious cause" (self-driving car) that has many SF detractors, they can get a higher rate of readers, which increases their impact (and maybe ad revenue?).
At least the number of articles suggesting that trolley problems are highly relevant to self-driving car implementations have gone down.
As a Mission local, I loved routinely seeing KitKat while strolling down 16th. He was friendly and made everyone’s day a little brighter. Perhaps other SF residents on HN feel the same.
Decades ago I accidentally ran over a cat while driving fairly slow on a residential street. It dashed out between parked cars and I barely saw it at all. I'm guessing sensor-laden driverless cars are actually going to turn out to be better at avoiding those and other kinds of accidents than human drivers.
Decades ago a kitty bolted in the middle of the street.
I was less than 18, using one of those little cars that reaches at most 50 Km/h.
I slammed the break and manage to stop maybe 2 cm from the kitty, which managed to continue out of the street alive.
The scooter behind me came close to me and complained that I almost killed them by slamming the breaks.
To this day, I still don't know if that was the right call. That guy could have been a dad and I could have killed a father.
Still I couldn't think of killing a cat either.
If a vehicle can't react to the emmergency break of the vehicle ahead, it's their fault. Change the kitty for a child and the complaint of the scooter is nonsense.
No. The other driver almost hurt himself following too closely, and not paying attention. There is nothing exceptional about emergency braking in a residential neighbourhood. Your prime responsibility when driving a car is in the direction of travel.
It sounds more like instead of learning a lesson about following too closely, he decided to turn the mix of anger or fear onto you. Hopefully, with time removed from the situation, he will realize that he should not follow as closely. And hopefully you won't be too affected by guilt-manipulation. (Obviously, it's a good idea when something behind you cannot slow quickly, to try not to brake too quickly, but in theory, a scooter should be able to stop very quickly.) With hard calls, you only can do your best with the information you had at the time.
Makes sense.
I did not veer because it's a very dangerous road and in Rome veering to dodge means you are hitting a scooter (scooters zip through traffic in Rome).
The reason why the guy behind me survived is because he did veer when I slammed the breaks
I remember vividly a childhood experience when a car I was in ran over a dog -- it ran towards the side of the car and went under the rear wheels of the car. I'm not sure there's any reaction time (human or otherwise) that would have prevented that from happening.
I had a dog bolt in front of my car once on a slow residential street. He survived. It was terrifying.
Still, this cat was on a busy stretch of 16th Street for nearly a decade and was unharmed by human drivers. I think Waymo failed pretty badly here. Some of the dismissive comments I've seen on this topic seem to me like they're making excuses.
Yep, I think it’s blatantly obvious that autonomous cars of some flavor will eclipse human ability sometime this century.
The interesting point will be when insurance companies reduce your rate if your car doesn’t have a steering wheel (or, equivalently, charge a “driving manually” fee). It might be obscured if car companies take on the risk themselves, but at some point people will start to notice that driving manually costs more.
I think they're close to it now - although there isn't enough data to drive home the proof yet.
Not "better than the best", but "safer than the average driver" - and if you aren't the only one on the road, your safety is a mix of your skill and everyone else's.
Given the limited data we have so far, it's undisputable that self-driving technologies that have been deployed commercially are dramatically safer than human driving. It will take a lot more data to know exactly how true this is, but in the meantime, 120 people die per day on average in the US due to traffic accidents.
> it's undisputable that self-driving technologies that have been deployed commercially are dramatically safer than human driving
This is really only true for Waymo, who appear to be the only folks operating at scale who did the work properly. Robotaxi, Cruise and all the others are in a separate bucket and should be statistically separated.
It's also true of Apollo in China (which has about as many miles logged as Waymo), and presumably, the limited operations of Zoox. I specifically referenced commercial operations.
Undisputable? Let's see what happens with the average "accidents per km" these firms keep touting once we let a bunch of self-driving cars drive on the ring around Paris or Antwerp.
So your only dispute is that you have no dispute, only the idea that things might change in the future, and some hypothetical dispute could emerge. Got it.
Seriously? You are claiming something is "undisputable" without citing any source or making any attempt at all to explain why that would be. I guess we really do live in a "post truth world" with people like you.
Maybe some of the cost savings from autonomous vehicles should be spent on separating roads from pedestrian walkways. I can imagine a world where roads are fully-enclosed in a fence and a segment gets shut down if an animal or human somehow finds their way inside (detected via computer vision).
If we're going to fully enclose and automate them anyway, we can probably have the cars going much faster and closer together, and save energy (and make it easier and safer to keep them that close together) by having cars going the same direction attach together. And the automation would be easier if we just had like different sets of cars on fixed tracks, with splits for when people need to go different ways. And at that point why do you even own your own car, they should probably be owned and operated by the government or a large public corporation. And...
Self-driving cars are constantly subject to mini-trolley problems. By training on human data, the robots learn values that are aligned with what humans value. -- Ashok Elluswamy (VP AI/Autopilot at Tesla)
If they were using my data I'd be partly responsible, due to failing to swerve around the last few suicidal prairie dogs I rolled over. I hate when that happens but I don't attempt high speed evasions. But I would if it were something larger, human or not, out of self defense. And it's never happened but I hope I'd stomp and swerve for a toddler. I'm happy with an autopilot learning that rule set, even though I've lost too many cats under tires.
You probably get more honest answers by presenting a trolley problem and then requiring a response within a second. It's a great implicit bias probe.
Interesting, I was in a (minor) accident with a waymo and a cat in LA.
The cat survived, but waymo had no idea about the cat. It definitely could see dogs on the sidewalk fine, but cat crossing the street is just too small to notice
I love waymo but I think people rushing to defend this are making a mistake. It should come at a high cost to the company, if they make any sort of tragic mistake like this such that they invest a lot to not commit it again.
Otherwise it's a slippery slope of "well but it's generally good"
"On Monday, a beloved shop cat was allegedly struck and killed by a Waymo driving down 16th Street. Now, a small sidewalk memorial has cropped up in his honor, complete with bouquets and lit candles."
Wonder why the title states allegedly but not the article?
Sad but at the end of the day Waymos are significantly less dangerous drivers than humans. If all cars on the street were Waymos, cats (and everyone else) would be much safer.
Why do you think cats would be safer? That's a confusing leap of logic. I think engineers tend to extrapolate from data, even in messy real world scenarios where the extrapolation doesn't make intuitive sense. It's an enormous social flaw that leads to bitterness between them and normal humans (autopilot vehicles are in fact the perfect storm that best exemplifies this problem).
Also note that there is an enormous issue of trust and dignity.
By "trust" I mean: We have seen how data and statistics are created. They are useful on average, but trusting them on very important, controversial topics, when they come from the private entity that stands to benefit from them, is an unrealistic ask for many normal humans.
By "dignity" I mean: Normal humans will not stand the indignity of their beloved community members, family, or pets being murdered by a robot designed by a bunch of techies chasing profit in silicon valley or wherever. Note that nowhere in that sentence did I say that the techies were negligent - they may have created the most responsible, reliable system possible under current technology. Too bad normal humans have no way of knowing if that's the case. Especially humans who are at all familiar with how all other software works and feels. It's a similar kind of hateful indignity and disgust to when the culpable party is a drunk driver, though qualitatively different. The nature of the cause of death matters a lot to people. If the robot is statistically safer, but when it kills my family it's because of a bug, people generally won't stand for that. But of course we don't know why exactly, as observers of an individual accident - maybe the situation was truly unavoidable and a human wouldn't have improved the outcome. The statistics don't matter to us in the moment when the death actually happens. Statistics don't tell us whether specifically our dead loved one would have died at the hands of a human driver - only that the chances are better on average.
Human nature is the hardest thing for engineers to relate to and account for.
It's not a leap to say that a driver that's safer to humans is also safer to cats. Human drivers try to avoid hitting humans and cats. Waymos make less driving mistakes in general. They're also never inebriated, tired, or inexperienced.
Legally, who is liable for a self-driving car which makes a mistake? Let's say it's egregious, and clearly the fault of the car, and it say kills someone?
In California, at least, Waymo is required to have $5 million in liability insurance. And the state has a law holding the manufacturer responsible in lieu of a driver. Though this setup has barely been tested since there have been so few incidents and the only "major" one (in CA) is still in court.
Humans drivers and self-driving companies that creates such hit-and-run situations should be prosecuted.
So the court can determine what to do next (jail, insurance, etc or just nothing). It does not matter if they hit an object, a pet or a child who didn’t look while crossing the street.
Perhaps assign a safety driver that puts its own driving license and criminal liability on the line, so the company cannot evade responsibility.
> Hitting a pet / animal should be treated the same as hitting a child? No
I think the point is you don't know for certain what you hit if you hit and run. The car should have enough collision detection to know when it's hit something.
That said, this story is sending up red flags with the "allegedly" in the title and lack of evidence beyond hearsay.
I mean, you can riff through the comments of some Mission-local Instagram posts about this incident. There are plenty of eyewitnesses, including someone who was behind the Waymo in question. I'm sure the "allegedly" is there for legal reasons.
Your safest bet is to protect as wide as possible the people and pets around you, today it is an object, tomorrow a pet, and eventually a child.
Pushing companies to investigate and take responsibility, and report these accidents is going to overall to improve reliability of the system.
The reality is that if you do not put strong punishments, these companies wont have the incentive to fix it, or they will push these priorities way lower on the to-do list.
