A company claims they have developed a fully autonomous all-weather plug-into-your-car system.
http://newatlas.com/aimotive-aidrive...
I'm really, really concerned that companies in the Silicon Valley tech bubble are pushing this tech too early.
They're not. From the article:
While Uber says the goal is full autonomy, the company admits the technology is not there yet.
California defines autonomous vehicles as cars that drive “without the active physical control or monitoring of a natural person.” Uber says that doesn’t apply to its self-driving cars, which can’t be driven without a human monitoring from the driver seat.
That said, Uber doesn’t expect to deploy fully autonomous, completely empty vehicles to pick up passengers anytime soon. The technology isn’t there yet, nor is the public ready to accept a vision of ghost vehicles roaming the streets at all hours.
This is another publicity stunt. I have no problem with it. There's a driver to take over and riders can opt out of receiving the ride. This is not fully self-driving tech, its a demo.
I have a lot of problems with pushing self-driving tech with a safety driver to take over, actually. The fact that you're depending on a driver who is more likely to be distracted to do the difficult bits makes the technology less safe.
Eh, the avg speed in San Francisco is like 10mph
YOU CAN HAIL A SELF-DRIVING UBER IN SAN FRANCISCO STARTING TODAY
For the last few months, Uber’s self-driving cars have been prowling the streets of San Francisco, forecasting the inevitable moment when the ride-hailing giant starts inviting passengers to take autonomous trips in the city where it first launched over seven years ago. That moment has finally arrived.Starting today, anyone in San Francisco who hails an UberX could find themselves in the backseat of a luxury, self-driving Volvo XC90, complete with leather interior, spinning LIDAR sensor, and a trunk full of computing power. It’s where I found myself last week, after being invited out to the Bay Area for a sneak peak before the official launch.
I am going there tomorrow! Going to see if I can grab one.
I have a lot of problems with pushing self-driving tech with a safety driver to take over, actually. The fact that you're depending on a driver who is more likely to be distracted to do the difficult bits makes the technology less safe.
There is ample evidence to support your concerns here.
There's even arguments that there's an inflection point with supervised automation where more automation becomes less safe, because you've already wrung out all the low-hanging safety gains from the automation, and all you're doing is making the supervisor less practiced by the time they have to take over.
The self-driving car that requires you to take over once every 2 years is a bloodbath waiting to happen.
Danny will pick you up! | Cancel
Ishmail will pick you up! | Cancel
Natasha will pick you up! | Cancel
Bob will pick you up! | Cancel
ROBOMASTER will pick you up! | YES!
The more I think about this, the more convinced I am that we have a problem on our hands.
My day job (aerospace engineer) is, ultimately, about transportation safety. My bread-and-butter is extremely low-probability events that kill people. The gold standard which we aim to mitigate any failure which could lead to fatalities is 1 event in every 1,000,000,000 hours of operation (1e-9). This results in an air travel environment that is, quite frankly, incredibly safe. You know the last time someone died because of commercial air travel in the US? 15 years ago. In New York. On September 11th.
Think about that for a moment. It is amazing. In the last year alone, there were over 9 million commercial airplane departures in the US, moving over 800 million people nearly a trillion miles. Not one of those people died as a result of those journeys. Zero. Zilch. Nada. And that applies for the preceding 14 years too. More if we don’t consider 9/11.
And yet, chances are that every single one of us knows someone that is terrified of flying.
This is the problem we have on our hands. One of the safest places you can possibly be, the cabin of a commercial flight in progress, scares the living bejeesus out of a significant portion of the population.
Even if we assume that self-driving cars can achieve a similar level of safety as aircraft (which is very much a best-case scenario – I expect that it’s a realistic long-term goal, but it’s going to take a while to get there), the sheer number of journeys involved means that we will see fatalities. In the short-term, these will be headline news in a way that routine traffic accidents aren’t today. They will occupy the national zeitgeist in precisely the opposite way we ignore the tens of thousands of traffic deaths we see annually now. Very few people refuse to travel in a car because it’s dangerous today (even though it is). Many likely will in the future (even though it won’t be).
Granted, this is likely a generational shift and something that will be short-lived, but I see it as a genuine threat to the widespread adoption of the technology.
Well if the media handles it responsibly.... well I should just stop talking there.
But seriously, events are inevitable and every time one happens they should throw up a stat showing how many traffic fatalities this year are the result of human error so far for context.
If self driving cars cause less than 30k deaths a year it's a win. Simple math.
In reality the number will be much lower.
Hopefully auto insurance will drive the tech and the adoption
Self driving car? $20 a month.
Drive yourself? $100 a month.
Or probably way more, proportional to accidents. Get people off the roads quick.
Well if the media handles it responsibly.... well I should just stop talking there.
Yeah they proved they value ratings and money above all else with the election this year
If self driving cars cause less than 30k deaths a year it's a win. Simple math.
Yes these kind of facts are always effective when dealing with people's irrational fears which is why no one is afraid of flying.
We've had this discussion before so I think this bears repeating: This is not a tech problem, it's a PR problem, and tech companies ignore massive looming PR problems at their own peril.
The more I think about this, the more convinced I am that we have a problem on our hands.
My day job (aerospace engineer) is, ultimately, about transportation safety. My bread-and-butter is extremely low-probability events that kill people. The gold standard which we aim to mitigate any failure which could lead to fatalities is 1 event in every 1,000,000,000 hours of operation (1e-9). This results in an air travel environment that is, quite frankly, incredibly safe. You know the last time someone died because of commercial air travel in the US? 15 years ago. In New York. On September 11th.
Think about that for a moment. It is amazing. In the last year alone, there were over 9 million commercial airplane departures in the US, moving over 800 million people nearly a trillion miles. Not one of those people died as a result of those journeys. Zero. Zilch. Nada. And that applies for the preceding 14 years too. More if we don’t consider 9/11.
And yet, chances are that every single one of us knows someone that is terrified of flying.
This is the problem we have on our hands. One of the safest places you can possibly be, the cabin of a commercial flight in progress, scares the living bejeesus out of a significant portion of the population.
Even if we assume that self-driving cars can achieve a similar level of safety as aircraft (which is very much a best-case scenario – I expect that it’s a realistic long-term goal, but it’s going to take a while to get there), the sheer number of journeys involved means that we will see fatalities. In the short-term, these will be headline news in a way that routine traffic accidents aren’t today. They will occupy the national zeitgeist in precisely the opposite way we ignore the tens of thousands of traffic deaths we see annually now. Very few people refuse to travel in a car because it’s dangerous today (even though it is). Many likely will in the future (even though it won’t be).
Granted, this is likely a generational shift and something that will be short-lived, but I see it as a genuine threat to the widespread adoption of the technology.
Does the zero deaths in 15 years include small aircraft? If not, then wouldn't that be more analogous to the automobile discussion than multi-billion dollar commercial aircraft? I can't afford a Tesla, let alone a Boeing product (well, maybe a t-shirt). I don't know a lot about small airplanes (or really any plane that doesn't carry weapons) so I don't know what level of automation they usually carry.
Is that stat of no deaths in commercial aviation not counting the crew of two that died in this flight? https://www.wikiwand.com/en/UPS_Airl...
Or this flight? https://www.wikiwand.com/en/Asiana_A...
Note: You're point still stands that it's super safe. I was just curious.
Does the zero deaths in 15 years include small aircraft? If not, then wouldn't that be more analogous to the automobile discussion than multi-billion dollar commercial aircraft? I can't afford a Tesla, let alone a Boeing product (well, maybe a t-shirt). I don't know a lot about small airplanes (or really any plane that doesn't carry weapons) so I don't know what level of automation they usually carry.
Am I including small aircraft? Most definitely not. There's plenty of small civil aviation fatalities.
I purposely excluded those, because to be honest, the level of technology in a self-driving car is a lot closer (by necessity) to a commercial airliner than, say, a Cessna. To be fair, you can go buy a small aircraft with fancy systems installed that are roughly analogous to something like a 737, but those are the kinds of business jets that fly rich CEOs around. The hobbyist private pilot, who owns a tenth share in a 30 year old Cessna (the likes of which make up the bulk of civil aviation) has likely zero flight-control automation.
More importantly, I expect the regulatory environment for self-driving cars to be a lot closer to that for commercial air travel than for civil aviation. That alone will drive some commonality between them, as a lot of the principles of safety-critical design (fail-safe/soft/tolerant systems and/or multiple layers of redundancy) leveraged to meet regulatory safety requirements in aerospace will be copied right over to the self-driving car world.
As to cost, you can make a million things a lot cheaper than if you're making a thousand things. A 787 costs you a quarter-billion dollars because Boeing needs to recoup the development cost over several thousand units. In 2015 alone, Toyota sold nearly half a million Camrys.
Is that stat of no deaths in commercial aviation not counting the crew of two that died in this flight? https://www.wikiwand.com/en/UPS_Airl...
Or this flight? https://www.wikiwand.com/en/Asiana_A...
Note: You're point still stands that it's super safe. I was just curious.
I was using NTSB data for passenger flights. The first link is a cargo flight with no passengers aboard, so won't show up in that data. There have been other fatal accidents with cargo airplanes in the time-span that I'm claiming superb safety record for.
Honestly not sure why the second one didn't show up in my research. It might be that the data is only tracking US-based airlines, for which I believe my assertion still holds true.
There's plenty of fatal accidents elsewhere in the world - the African continent, for instance, is a hotbed of fatal air crashes, usually involving very old, poorly maintained aircraft that wouldn't be legally allowed to fly in FAA-controlled airspace.
I think comparing self-driving car safety to airline travel safety is a bad comparison. First, you're talking a couple orders of magnitude difference in the potential fatalities from a single incident. Second, cars as taxis in cities (which is what Uber is doing) have a lot less potential for damage than an airplane falling out of the sky onto a city. Taxis typically travel at slow enough speeds that even the worst collision would not seriously injure a passenger, given the safety features of modern cars. Pedestrians are another matter, though. There's more traffic around taxis, but that's something a computer is good at dealing with. In fact, the more traffic, the better the computer is than a human.
I think Stele has it right. If the number of incidents is demonstrably less than human drivers, it's a win. I don't think the companies renting out their self-driving cars will have to deal with the kind of fear the airlines deal with. A car is a much more familiar piece of equipment. How often do most people take a ride in an airplane? Compare that with how often most people ride or even drive a car. Driving a car is something that is familiar and that most adults understand. Driving an airplane is a feat of wizardry to most adults. Finally, cars roll around on one of humankind's oldest inventions, the wheel. Airplanes stay aloft using fairy dust and blood sacrifices.
I definitely agree that the less a supervisory human must take over, the greater the hazard. However, I suspect that in reality, the "drivers" are only there to reassure the passengers that the crazed human-murdering AI in the trunk will not drive them off a cliff.
I think comparing self-driving car safety to airline travel safety is a bad comparison. First, you're talking a couple orders of magnitude difference in the potential fatalities from a single incident. Second, cars as taxis in cities (which is what Uber is doing) have a lot less potential for damage than an airplane falling out of the sky onto a city. Taxis typically travel at slow enough speeds that even the worst collision would not seriously injure a passenger, given the safety features of modern cars. Pedestrians are another matter, though. There's more traffic around taxis, but that's something a computer is good at dealing with. In fact, the more traffic, the better the computer is than a human.
You’re not wrong, but I disagree. In terms of the regulatory environment (which provides the overriding set of design requirements/constraints when it comes to safety), there is no difference between an aircraft accident in which one person dies, and one in which everyone onboard dies with further fatalities on the ground. They are both considered equally unacceptable and mandated to be extremely remote. If we assume a similar regulatory environment, the comparison holds. Though to be fair, that assumption may itself be flawed.
One interesting wrinkle to consider that makes the self-driving car *more* dangerous is its proximity to uninvolved people. To whit, it’s rare for anyone to be killed in an airplane accident who hadn’t explicitly chosen to be onboard that aircraft (the vast majority of the world’s surface is either ocean or uninhabited/lightly inhabited. Plane crashes in the ocean or a field don’t kill anyone who isn’t on the plane). Pedestrians are a thorny topic there – it’s more difficult to argue that they put themselves in harms’ way simply by virtue of crossing the road.
I don't think the companies renting out their self-driving cars will have to deal with the kind of fear the airlines deal with. A car is a much more familiar piece of equipment. How often do most people take a ride in an airplane? Compare that with how often most people ride or even drive a car. Driving a car is something that is familiar and that most adults understand. Driving an airplane is a feat of wizardry to most adults.
I can’t argue with any of that, but to provide a counterpoint, how common is being a nervous passenger in a car? Now, what if the other driver is a literal feat of technological wizardry?
It seems Google spawned off their self-driving car project into its own company (under Alphabet?) with a goal of commercializing their cute control free cars.
Blah! I don't need more information on the car. I need the car!
It'll cost Waymo' than I can afford....
Ok, I'm done. =)
Say hello to Waymo.
No! I refuse! YOU CAN'T MAKE ME! YOU'RE NOT MY REAL DAD!
Uber's self-driving car launch in San Francisco has a problem: California says they didn't get the required permits.
Uber Launches Self-Driving Cars In California, Over State's Objections
Uber did everything right in Pittsburgh with its self-driving cars — but is doing everything wrong in San Francisco
(Also, their employees have been spying on people: Uber Whistleblower Says Employees Used Company Systems to Stalk Exes and Celebs)
For the regulatory environment, I'm going to be watching with interest because I've had the chance in the past to interact with some of the people and agencies doing the regulation. Watching the process is going to be fascinating.
For the US at present, the DOT takes a much more lax approach to ground vehicles than it does to aviation. This is in sharp contrast to, say, Sweden. Which has the explicit goal of attempting to eliminate all road deaths. They're well on their way to that, with deaths down to 3 per 100,000 (compared to the US's 11.4).
Volvo, as has been mentioned before, accepts all liability when their cars are in autonomous mode. If US manufacturers truly get on board with that, car safety will be much more comparable to airplane safety. Given their past safety record (where they spent decades trying to suppress the seat belt and airbags to save money) that is a bit of a long shot, but they might get dragged into it despite themselves.
I will say that IIHS, at least, is ahead of the curve. A few months ago I got to see their new testing facility for autonomous cars. They're putting a lot of effort into figuring out what kind of tests the things are going to require.
For the US at present, the DOT takes a much more lax approach to ground vehicles than it does to aviation. This is in sharp contrast to, say, Sweden. Which has the explicit goal of attempting to eliminate all road deaths. They're well on their way to that, with deaths down to 3 per 100,000 (compared to the US's 11.4).
I have a lot of thoughts on/problems with Uber and the regulatory environment, but for now I just wanted to pop in and say that the USDOT and Federal Highway Association have launched a Towards Zero Deaths program, emulating the Sweden program. Many other state and local groups have also committed to this goal, though it is a much more recent thing here.
Slate considers a consequence of self-driving cars;
Decreased availability of viable organs for those in need due to fewer people dying in car accidents. Then again most evidence suggests we could boosts organ donation rates to near 90% by simply changing the wording on the card to "check this if you do NOT want to donate" {as tested and verified as successful in many European countries, good ol choice paralysis} and we can't be bothered to do that so I have to think we're not really that concerned about organ donation, sadly.
Wait... so we WANT people to keep dying in car crashes? So that we can... harvest their organs to save people that are almost dying in car crashes?
Ugh... How about instead of counting on death by mutilation, we actually get off our asses and get organized about stem-cell and cloning research and start growing organs in a vat? Enough organs for everybody, possibly even 3D printed from your own tissue (i.e. no rejection once implanted).
[the rest of my comment will remain silent for fear of taking this to Cleveland]
Slate considers a consequence of self-driving cars;
Decreased availability of viable organs for those in need due to fewer people dying in car accidents. Then again most evidence suggests we could boosts organ donation rates to near 90% by simply changing the wording on the card to "check this if you do NOT want to donate" {as tested and verified as successful in many European countries, good ol choice paralysis} and we can't be bothered to do that so I have to think we're not really that concerned about organ donation, sadly.
Yeah. I'm going to file that under satire because it sounds a whole lot like the sort of reasoning in swift's modest proposal.
krev82 wrote:Slate considers a consequence of self-driving cars;
Decreased availability of viable organs for those in need due to fewer people dying in car accidents. Then again most evidence suggests we could boosts organ donation rates to near 90% by simply changing the wording on the card to "check this if you do NOT want to donate" {as tested and verified as successful in many European countries, good ol choice paralysis} and we can't be bothered to do that so I have to think we're not really that concerned about organ donation, sadly.
Yeah. I'm going to file that under satire because it sounds a whole lot like the sort of reasoning in swift's modest proposal.
The basic argument of the article is that we need to "prepare" for the future, not prevent it.
Also, the article kind of overstates it's point as the article itself says that only 20% or organs come from car crashes so even if we get that down to 0% it's still "only" a 20% reduction. Bad, but not "radically change the organ donation market"-bad.
Side note: If you have an iphone it takes about 3 minutes to sign up for organ donation under your "HEALTH" app. Hopefully you are signed up in your drivers license, but that doesn't always carry across state lines whereass Apple's partnership with Donate Life America apparently does. I encourge those of you who have an iPhone to take the five minutes;
Sign up to be an organ donor
In the United States, you can sign up to be an organ donor in the Health app under Medical ID. Follow these steps:
Open Health and tap Medical ID.
Under Organ Donation, tap Sign Up with Donate Life.
Fill out the registration form, then tap Continue at the bottom of the screen.
Confirm your registration, then tap Complete Registration with Donate Life.
You can edit your information with Donate Life at any time. Just open the Health app, tap Medical ID, and tap Edit under Organ Donation Registry. To withdraw your registration, scroll to the bottom of the Edit screen and tap Remove Me.
For information on Donate Life America’s privacy policy, please visit their website.
Pages