Slashdot Mirror
Fully Driverless Waymo Taxis Are Due Out This Year, Alarming Critics (arstechnica.com)
Alphabet's Waymo is launching a driverless taxi service in Phoenix in the next three months -- and it's open to the public. But due to the limited regulations surrounding self-driving cars, many critics argue that more regulations are needed to ensure the safety of these vehicles before they roll out for public and commercial use. Ars Technica reports: If a company wants to sell a new airplane or medical device, it must undergo an extensive process to prove to federal regulators that it's safe. Currently, there's no comparable requirement for self-driving cars. Federal and state laws allow Waymo to introduce fully self-driving cars onto public streets in Arizona without any formal approval process. That's not an oversight. It represents a bipartisan consensus in Washington that strict regulation of self-driving cars would do more harm than good.
Mary "Missy" Cummings, an engineering professor at Duke, agrees. "I don't think there should be any driverless cars on the road," she tells Ars. "I think it's unconscionable that no one is stipulating that testing needs to be done before they're put on the road." But so far these advocates' demands have fallen on deaf ears. Partly that's because federal regulators don't want to slow the introduction of a technology that could save a lot of lives in the long run. Partly it's because they believe that liability concerns give companies a strong enough incentive to behave responsibly. And partly it's because no one is sure how to regulate self-driving cars effectively. When it comes to driverless cars, "there's no consensus on what it means to be safe or how we go about proving that," says Bryant Walker Smith, a legal scholar at the University of South Carolina.
Mary "Missy" Cummings, an engineering professor at Duke, agrees. "I don't think there should be any driverless cars on the road," she tells Ars. "I think it's unconscionable that no one is stipulating that testing needs to be done before they're put on the road." But so far these advocates' demands have fallen on deaf ears. Partly that's because federal regulators don't want to slow the introduction of a technology that could save a lot of lives in the long run. Partly it's because they believe that liability concerns give companies a strong enough incentive to behave responsibly. And partly it's because no one is sure how to regulate self-driving cars effectively. When it comes to driverless cars, "there's no consensus on what it means to be safe or how we go about proving that," says Bryant Walker Smith, a legal scholar at the University of South Carolina.
"Currently, there's no comparable requirement for self-driving cars."
Human drivers cause 6,5 million accidents per year, killing tens of thousands of people and injuring several millions.
This can only be better.
Mary "Missy" Cummings, an engineering professor at Duke, agrees. "I don't think there should be any driverless cars on the road," she tells Ars. "I think it's unconscionable that no one is stipulating that testing needs to be done before they're put on the road."
What does she assume this whole time self driving cars have just been something in people's heads? Every company who's in on this technology brags about their logged road time... Glad she ain't my prof.
I tend to rant.
I don't worry much about the cars themselves, but about the potential for legislation mandating we have self-driving cars. If you are old enough, you'll recall all the arguments about motorcycle helmet laws. How motorcycle accidents without helmets overloaded ER rooms and cost the public a lot of money in medical care. There is a clear precedent for my fears, The argument is the same. Don't forget that driving is a privilege not a right.
As previous accidents have shown, a human is a lousy backup for a self-driving car. Very few humans, if there exists any at all, can maintain constant vigilance over a system that does what its supposed to do 99.999% of the time.
We don't understand self-driving tech and their failure modes well enough to create the equivalent of a driving test. People will die before that can be developed, and the faster and more frequently they die, the faster the development. The only alternative is to let someone else develop it first and buy the tech from them.
Fine for a family car with self-drive (or even Tesla's driver assistance) functionality, but we're specifically talking about self-driving taxis here. What is the humans in the taxi are using it because they are not able (legally and/or physically) to drive, and that's why they are taking the taxi in the first place? It's going to be kind of limiting for a taxi if it can only accept passengers that are able to drive it as well, especially on a Friday or Saturday night, are they also going to have to require the designated driver to scan their driver's license and to breathe into a breathalyzer before setting off?
I guess there would be slightly less chance of the floor getting covered in vomit though, so there is that.
UNIX? They're not even circumcised! Savages!
I used to work for a medical device manufacturer. While having to deal with a lot of regulations was certainly annoying (mostly because they are written by lawyers and you need to be a lawyer to really understand them), the great thing about them was that once you complied, you didn't have to worry nearly as much about liability. If there were no regulations (basically a form of self-regulation), then how exactly do you prove that you were not negligent? Maybe you think all the tests you did were enough. Maybe the lawyers you hired for advice thought so too. But you'll never know until it is tested in court. With regulations it is more-or-less black and white as to whether you have done enough to absolve yourself of responsibility for unforeseen events.
Another important point is that regulation creates a powerful barrier to entry in a market. The infrastructure required (in terms of processes and procedures) is immense, and large companies can gain economies of scale for these work. While the tech is enough of a barrier to entry right now, as time goes on this will change for driverless vehicles as well.
All they need to do is kill one pedestrian, and then it will be the same shit.
How much adversarial attacks have those cars been exposed to? Just because a car can safely drive down a standard road while supervised, doesn't mean it can't fail in catastrophic ways when exposed to non-standard situations. You don't want to have some jocksters paint stripes down a cliff and the cars blindly driving to their doom Wile E. Coyote style.
I wouldn't trust those cars one bit until they have been shown to be able to handle freak situations in a reasonable way.
I wouldn't trust those cars one bit until they have been shown to be able to handle freak situations in a reasonable way.
That statement should apply to both self-driving and human cars: No human-driven cars should be allowed on the road until humans have been shown to handle freak situations in a reasonable way. Sadly, this is provably not the case.
They have, the difference is these won't have a backup driver to take over.
It seems to me, that Waymo have the best tech and take the safety aspect very seriously, so I don't see any cause for alarm myself.
If Uber were doing this, then there would be cause for alarm. Regulation is needed, as although Waymo take safety seriously, some of there competitors may not.
I agree completely. As over-the-top safety conscious as Waymo has been taking baby steps all the way to release- if Waymo think they're ready I'm inclined to believe them... but like you, I wouldn't believe some of their competitors.
If Waymo are being premature with this and any harm comes to anyone; you can bet they will get their socks sued off them.
"That's the way to do it" - Punch
It is inevitable that they pass laws allowing machines to kill x number of people. It can be no other way. And that will be a major devaluing of human life.
The question is, will it be more or less devaluing than the currently allowed rate of 40K+ people a year killed by human-driven cars in the US?
The fact that you think people function like computers is amusing... and revealing.
Well, Waymo certainly has the best PR.
They also only have 9 million miles on the road, most of that effectively in a "sandbox". In the US, there's one fatality every 86 million miles. So the fact that Waymo hasn't killed anyone yet is hardly indicative of anything.
Leaving it to corporations to regulate themselves due to fears of liability has created one disaster after the next. And come on, let's not act like we can't all figure out how autonomous vehicles could be tested. Give a small fleet of them to the NHTSA and have the NHTSA spend a few weeks subjecting them to one unanticipated event after the next in an (easily reconfigurable) mock town, without a given "script" that the manufacturer could use as a cheat sheet.
Too onerous of a testing cost? At least give them a one day serial battery of scenario tests. I mean, come on. We're talking about peoples' lives here. It's bad enough that Level 2 systems don't have to do this. But Level 5? Ugh.
"Close the door! What, were you born in a barn?" -- Police chief, "Jesus Christ Supercop"
Sure, the FATALITY rate is roughly 1.25 per 100,000,000 miles in the US, but the ACCIDENT rate is around 600 per 100,000,000 miles.
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
That's bull, because if a human driver kills a human, then that human has made a mistake. If an automated car kills a human, THEY ALL make the mistake.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
Governments have safety in mind, people have career progress and short term stock prices in mind.
Waymo cars required an intervention every 5600 miles, the average American driver drives 13000 miles a year. The average human doesn't crash 3 times a year.
I wouldn't trust those cars one bit until they have been shown to be able to handle freak situations in a reasonable way.
That statement should apply to both self-driving and human cars: No human-driven cars should be allowed on the road until humans have been shown to handle freak situations in a reasonable way. Sadly, this is provably not the case.
Humans are a few orders of magnitude better than the numbers that waymo posted for there cars. Typical humans go 100s of thousands of miles without needing an intervention. Waymo cars (according to their own data) need an intervention every 5600m or so.
So, yeah, humans may not be perfect, but according to the numbers they're a hell of a lot better than the best SDCs available so far.
I'm a minority race. Save your vitriol for white people.
Without knowing the circumstances of each "disengagement", I think it is presumptuous to assume they would have all led to accidents.
It could be analogous to a human driver pulling off the side of the road in heavy rain - a fail safe.
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
And then THEY ALL can be fixed, probably with an over-the-air update. No drunk driving PR campaign, no driver outreach or training, no expensive modification of signage or intersections, etc. This is an advantage. Only a hardware limitation would be similar to the current paradigm, and even then it would be no worse.
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
... Their idea: "Just ride bikes. If you live too far out, move in closer." ...The snotty, "just ride bikes" suggestion sounds great, but in reality, it is pompous. There are many elderly people or people in wheelchairs that can't go and buy an Orbea Orca and magically be on their way.
In reality, I could ride my bike to the grocery store if just picking up a few items. I can't ride my bike to work, it would take too long (and I can't afford to move closer to work)... but there are a few other locations within a few dozen miles that I could probably use my bike for... but I won't.
Two reasons:
1) It's friggin' hot in the South 6 months out of the year... I'd arrive everywhere stinking awful. Not to mention, can you imagine all the dehydration deaths if many people did this?
2) It's not safe. I see more "white bicycle memorials" marking where cyclists are killed than I see actual bicycles on the road around here. People don't drive safely around cyclists here and there are no bike lanes. I know this would change as people got more experience around cyclists and such... but I wouldn't want to be cycling until it is the norm.
"That's the way to do it" - Punch
Also true of signage, intersection design, human training, etc.
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
If that is your threshold, then just go away unconvinced because that kind of analysis isn't going to happen in a slashdot discussion thread. Comparing the systemic complexity of transportation system design with and without the added variable of artificial intelligence is thesis material, at least. All you'll get out of me is that they are both very hard, subject to continuous change and improvement, and both very imperfect. No, I don't know the relative magnitudes and neither do you.
W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
Waymo used to publish a monthly list of all accidents its cars were in, but they've since stopped doing this in january 2017 and scrubbed their website of past lists. And the claim that it's not been at fault in accidents when under its own control are not true - for example...
It's important to remember that for most of its miles, it's also had a human present who can take control to prevent accidents, and a large chunk of its miles have been "sandboxed" - that is, Google/Waymo tightly controlled where it was allowed to drive, in what conditions it was allowed to drive.
"Close the door! What, were you born in a barn?" -- Police chief, "Jesus Christ Supercop"
No, this is not accurate.
The Waymo vehicles have been driving around Phoenix, without a driver, for most of this year. The vehicles have been limited to 'beta testers' who could hail a ride. This announcement is really just about the commercial launch of the service. At that point everyone will be able to hail one.
Also the notion that there is no 'backup driver' is false. The driver is just not in the car. If the car gets stuck it will simply stop, turn on the hazard lights, and wait for a remote operator to do something to help.
I might add, it's not just 3.22 trillion miles driven by humans per year, but those miles are anywhere in the continental US, in any weather and any situation; not just on clearly marked roads with proper traffic signals.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
Humans drive 3.22 trillion miles a year and get into around 20 million accidents (yes I have accounted for 30% unreported accidents). This means they drive 161,000 safe miles per accident, giving them a failure rate of 1 every 161,000 miles. Let's not quibble about the fact that this success rate is beyond comparable to self driving because this is all weather, all places, all circumstances. You can't tell me Apple has less than a 0.0006% failure rate for their electronics, or that people get into accidents more often than vehicles break down.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
The Tesla that killed the dude by driving straight into a crete highway barrier had just gotten an update. The previous version wouldn't have wrecked like that.
That was just lane following, a much simpler problem than general autonomous driving.
John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
Actually, the question I can't help wondering is whether unused deaths will be able to be traded like carbon credits.
When someone says, "Any fool can see