You're all too skeptical of super-duper self-driving cars, apparently Lamenting the 35,000 people in the US who died in motor vehicle crashes in 2015, Senator John Thune (R-SD), chairman of the Senate Committee on Commerce, Science, and Transportation, on Wednesday held a hearing in Washington, DC, on the need to win public support for automated cars and to rethink regulation. "Industry must …
I'm in charge of my life, my choices - good or bad are mine. I live and die by them.
I will not allow a machine to be in charge of something that is life or death for me or others around me.
I've seen plenty of badly written software and until you can demonstrate over a long period that the software is 100% error free then I might change my mind.
Of course, when the inevitable bugs do come out from the edge cases that were never tested by the manufacturer, they need to be liable with big, really big fines, otherwise, its just a different tax to get sales of their latest toys that will have been developed in the normal corporate manner of the least cost to the business with the minimally complaint solution and cheapest components they could source.
Perhaps if you started by taking the finance people out of the loop, put the engineers back in control and doing things right then I might consider it in a couple of decades.
Until then, the answer is no.
Oh and I quite enjoy being in control of my car, its a relaxing past time and its part of what makes me unique - the choice of model, style, colour, what I put in it, how I drive it, where and when I drive it, etc.
I'm not so sure that I'll enjoy cleaning it and polishing it when its jut a computer controlled gadget - just like I don't enjoy cleaning the house or cleaning the dust out of my computer when its getting hot and starts to do odd things. Imagine if a car, built with the same basic lego bricks were to suffer the same sort of problem.
Is anyone else wondering what happens if you put a million cars on the road with different laser and lidar systems on them and they all start receiving each others signals or blinding every other living thing withing a 300 metre radius.
"I must be strange then as I really like cleaning my pc and even other peoples."
If you can imagine it, then there's a specialist fetish site on t'interweb where you can get yer rocks off!
NB, all links are SFW El Reg pages :-)
Luddites will get sorted by insurance costs for human driven cars going through the roof, once self driving cars have proven a superior safety record. You sound like one of the people who resisted antilock brakes, believing you could do a better job of braking than a computer.
I don't agree with a rush forward into self driving cars, and I don't think they'll come as fast as some people seem to think (talk about 2020 is ridiculous) However once testing can prove a lower fatality & accident rate than human driven cars, I would be fully in favor of the government using laws to encourage their rollout and discourage easily distracted meatbags behind the wheel.
As for cars with different systems interacting....yeah, I'm sure no one designing them has thought of that! :P
"As for cars with different systems interacting....yeah, I'm sure no one designing them has thought of that!"
Sadly, that's an entirely likely outcome. The car companies, like Ford, VAG etc. don't have the experience in software engineering at this level (think IoT levels of security) and the software companies, like Waymo, Uber etc., don't have the structured design and test ethos of the car companies building to regulated safety standards.
@DougS: "Luddites will get sorted by insurance costs for human driven cars going through the roof, once self driving cars have proven a superior safety record."
How does that make sense? Is a human driver surrounded by super-safe robot cars, more likely to have an accident than the same human surrounded by other cars driven by unreliable humans? No. If the robot cars really are safer, then they'll be better at avoiding accidents and will reduce insurance costs for the remaining human drivers. Not to the very low level that the owners of the robot cars would be paying, but lower than now.
@DougS: "Luddites will get sorted by insurance costs for human driven cars going through the roof, once self driving cars have proven a superior safety record."
@I am the liquor: "How does that make sense?"
There are a couple influencing factors. First, as the number of people who are self driving goes down, so does the amount of money that insurance companies will be collecting as premiums. After all, the robots will be safer and people will expect that the cost of insuring robot drivers will be lower than insuring meat sacks.
Second, the cost of cars will go up as more tech is thrown in them. So insurance companies will have to squeeze the meat sacks more. Sure you might be the same driver you were but if the vehicles are more expensive and robot drivers are safer then by definition the amount of risk you represent increases - hence higher premiums.
Bear in mind that there are plenty of accidents that can occur when only 1 vehicle is in motion and if that 1 driver is a meat sack which is easily distractable then...
@Bob Dole (tm):
You seem to have the same misconception that DougS had. Yes, as the number of people who are self driving goes down, so does the amount of money that insurance companies will be collecting as premiums. But as the number of people who are self driving goes down, so does the number of accident payouts that the premiums have to cover.
Your insurance premium is (the probability that you will have an accident) * (the cost of that accident) + (the insurer's profit). The probability that you will have an accident doesn't increase if all the other cars on the road become robotised. In fact, it goes down. So does your premium. Robot drivers becoming safer doesn't make you more risky.
You contend that the costs of accidents will increase because cars will become much more expensive: I'm unconvinced. Today's cars contain vastly more technology than those of decades ago, but in real terms the prices are not much different.
Who is going to insist on doing their own driving? The best drivers? Unlikely. Also consider what happens in a world with 95% self driving cars to the 5% who insist on doing their own driving, and they have an accident that is their fault and they kill or seriously injure someone in a self driving car. Today 35,000 people die in auto accidents every year in the US - enough that while it is always a tragedy, people accept that while the risk can be minimized it can't be avoided.
In the world with 95% self driving cars with a much better accident record, juries will really stick it to those who choose to do their own driving and have a bad accident that a self driving car would not have. Instead of being sued for a million bucks like today, you'll be sued for 10 million. They'll say "this could have been avoided if this person had just let the car drive him like a normal person". Premiums will need to reflect much higher liability limits that will be required for those choosing to drive themselves.
Go price a policy with 10 million in liability if you think its cheap. I have an auto policy with $250K or something like that of liability, and a $2 million umbrella. I pay $250/yr for that umbrella - the difference between the $250K and $2 million in liability (it also covers me for my home, but it is pretty hard for someone to slip on the ice and sue for more than $250K) I expect it would be at least $1000 more if I wanted 10 million in liability.
Sounds like things are very different in the US. Juries in civil tort cases? I bet the lawyers love that. Juries are mostly for criminal trials in the UK. Also there are no punitive damages. If you did £1m of damage then you get sued for £1m, you don't get sued for £10m as extra punishment for your bad life choices.
Standard car insurance policies in the UK have a £20m limit on 3rd party liability for property damage, and no limit on 3rd party liability for personal injury.
On your first point, I think it's both wrong and irrelevant. Irrelevant, because it doesn't affect the risk profile of an individual driver. So what if all the remaining human drivers are bad drivers? They were bad drivers before too. Their risk of having an accident has not increased because of all the good drivers switching to robot cars. Wrong, because I'd expect the drunks and the careless drivers who natter on their cellphone are the most likely to prefer robot cars.
"I will not allow a machine to be in charge of something that is life or death for me or others around me."
So, no pacemakers, airplane autopilots, traffic control systems, train control systems, water, sewage, or power control systems, etc., for you? I've got some bad news for you about the many, many things which have been operated by computer control for, you know, decades.
Pacemakers - no
Autopilots - pilots still control and have overrides, they also go into simulators regularly to keep the skills so they can try and recover when something goes wrong, it's not the computer that is expected to fix a problem, since there is no camera to tell it that a piece of the wing came off or that a lithium battery just filled the plane with some bad smelling smoke or any of a stack of other potential failure scenarios that the computer has no sensors for or code to interpret that special case that it causes on the sensors it has got.
Take that training need to self driving cars - will people be expected to attend regular driving lessons and will the manufacturers provide simulators - no, they will just claim it's not necessary at all
Traffic control systems - they advise us, but we still make the decision if we trust the info it's giving us as we are accountable for our actions. Crossing lights out, take extra care and work out when to cross. What would an autonomous car do - fail to sense it or wait till it comes back on sometime next Tuesday.
Train control - still has a meat bag at the front, at least on most U.K. Trains. Ok, some of them do get it wrong, but they generally get my vote.
Power control systems - I'm not plugged into the mains, so if it goes off then it's just inconvenient and I'll carry on. If the power goes up and it all starts going pop, then I can walk away to safety until a human fixes things.
The trend is that I'm in control and am accountable for me, not a computer where I'm expected to put my trust in a piece of code written by the cheapest programmer they could get who doesn't have any accountability for their code.
Insurance?
Surely the makers of the automatic cars will be the ones who need to insure them?
You insure your car in the event you break it. If [CAR COMPANY AI] is the one driving it, surely [CAR COMPANY] should be liable if the driver wrecks it?
You might have to have a passive insurance for such things as doors in parking lots, weather events, keying etc. - but accidents while driving must surely be the responsibility of the driver - that is [CAR COMPANY]?!
I take it you don't fly then? Or use public transportation in any shape or manner? There are times when we don't have control of our destiny. Even vehicles we drive ourselves can have issues beyond our control like throttles suddenly going wide open, etc. These things have consequences not just for ourselves but others.
However, I do appreciate the sentiment. I'd rather do the driving of my car.
You set an impossibly high bar. Automated cars will never be 100% error free.
There are 35,000 road deaths in America every year. If automated cars could even halve that number would it not be wise to mandate the use of automated cars? How many human lives is driving autonomy worth?
The way to mandate robot car usage is likely, I think, to be the same way they're controlling traffic in city centres; first a high access charge for non-robot cars and then eventually a ban on them. Roads which are 'harder' for the robots will get controlled last and remain available for those who actually *like* driving.
As an aside, and a response to the point upthread about a million cars all shining lights at each other... I've wondered - on no evidence whatsoever - whether current methods that use all sorts of sensors to try and build up a picture of the surrounding environment are necessarily the best way to do it, and whether perhaps we might observe that people manage to control a car with nothing more than two eyes, mostly facing forwards only?
Sure, people control a car using two eyes facing forward. That's why we sometimes have accidents from cars in our blind spot, or around a curve at night, or in foggy conditions. It makes sense to use more than one method to "see", no need to emulate us limited humans and give a car a single pair of eyes on a rotating gymbal that has to decide which way to look.
All of them.
Seriously I am sick of the 'no risk is allowed' culture, I am sick of the 'can't do anything in case someone stubs their toes' and sick to death of being dictated to.
I would like a self driving car for the occasions when the distance is great, I am tired, I want to talk to the kids etc etc but I also want a car I can take the pleasure of driving. Its like I want to go to the supermarket and buy some tomatoes, but I also want to be able to grow them myself and go into the garden to eat them!
In charge... oh yeah... you never fly, never sail on a ship, never use a train? You control what the other idiot drivers on the road are doing? You control the earthquake and the standard of the building you are in? We all take risks all the time.
The question is which is the better risk/fun equation. I would quite happily let a car drive the 5 or 6 hundred miles of motorway when I go home if I was tired and could sleep instead of dying in a car crash. On the other hand on a nice sunny morning when wide awake and on a fun curvy interesting road with few cars and some nice scenery, window down.... then I want to be in control
(Or your own choice of game)
While these AI can do wonders, they still fail. They may fail less than humans do. But the problem is not lack of confidence. I'm confident that we can all have flying surf boards (they are buildable, just noisy, people are doing it, it's just a model aircraft under your feet :P ). I'm not confident they won't drop from the sky when someone forgets they only have 7 mins worth of fuel.
Likewise I don't think people have a problem with automated freeways/motorways or automated parking etc. But the drive to "automating all the things" seems steeped more in marketing speech, than in reality or the laws of physics.
Every time I shudder at the idea of fully automated cars, I consider how really bad humans are at driving, at calculating risk, and at evaluating ourselves.
If the worst automated car is better than the average teen with 2 months' experience, that's a step up. If it learns from mistakes, that's already better than the average human.
Risk from cyber attacks? Is that risk worse than a human who uses the Internet while driving?
If nothing else, automated cars permit objective regulation of driving behavior, something which we humans are unable to do.
We could learn from aviation, which uses automation very effectively. No, not fully automated driving, but Instrument Landing Systems, TCAS collision avoidance, GPWS warnings about terrain, FADEC engine controls, Fly By Wire which compensates for mechanical control problems, and a regulatory environment that requires manufacturers to prove the safety of their products.
Oh, and aviation considers any "accident" as a failure to be investigated and hopefully prevented in the future. In an automobile, if you drive on bald tires and then slide your car into oncoming traffic, that's an unforeseeable act of God. Car culture is no paragon of safety.
"If the worst automated car is better than the average teen with 2 months' experience, that's a step up. If it learns from mistakes, that's already better than the average human."
Let's think that one through. You choose an "average teen with 2 months' experience" as you yardstick. Why not a 40 year old with over 20 years of driving experience? Presumably because the latter is a much better driver and wouldn't be such an easy target for the automated car to beat (and don't think I didn't spot that weasel "if" at the start of the sentence).
So it rather looks as if you think humans can learn from something, whether it be from mistakes or from simple practice. And yet you say they can't.
Okay, how about the 70-year old with failing eyesight, reflexes, and mental capacity, who is a danger to self and others but refuses to admit it and is only going to get worse? One of the main arguments for allowing senior citizens to keep their licenses despite degrading faculties is the loss of dignity and independence that would ensue. What if, rather, those older folks had a dignified answer in the form, in essence, of a robot chauffeur, would that really be so bad?
"Why not a 40 year old with over 20 years of driving experience? Presumably because the latter is a much better driver
I've met plenty of 40+ year olds with decades of driving experience who are truly frightening to ride with. The issue is not so much to do with whether or not humans can or can't learn from their mistakes, but rather how much damage and/or loss of life can be mitigated by automated systems.
Even the very best drivers can make mistakes due to fatigue, distraction or a host of other events. Automated systems have the advantage of never getting tired, and never losing focus on the task at hand. This has already been well and truly proven in areas like manufacturing, to the point where automation is recognized as a key method of removing risk of injury in many workplaces worldwide.
Lastly, keep in mind that while most road rules seem restrictive, they exist to protect the masses from the lowest common denominators. No matter how skilled you may be, there's always a chance that someone else's inattention might lead to your death.
Meph, indeed. However how will the clever car handle the willfully dangerous ? Think of drunks, druggies and the inevitable clown in a BMW/Merc/light truck with a severe ego problem ? IMHO, skepticism of the merchants is warranted. The best way to sell driving automation is already slowly happening by implementing driver assistance. Cruise control, automatic collision detection and lane drift prevention. As these become standard and extended to, for instance, something that prevents compulsive tailgaters, user acceptance will rise as well as vehicle manufacturers and their coders getting improved edge case management. Incremental improvements over a decade or two may also allow laws to be adapted appropriately insteady of panic driven. No pun in 10d. In long term a mostly self driving car is a great goal. Whether it is currently practical I believe debatable.
Comparison with aviation autopilots in cars is a category confusion. Airfields have the same design, (mostly flat) and fixed wing aircraft have similar characteristics to control for. Also as some-one else has commented, civilian aircraft rarely come close by design and the only ones that do, gliders, have pilots trained in this type of aviation and are not transport vehicles in built up areas. Vehicles are the opposite, close together and often very different speeds.
Yea? What happens when someone hacks the systems that run automated cars. Not just your car or his or hers, but several? How is that safer than humans driving their own cars? There's no scenario where a bunch of drivers all of a sudden decide to drive badly. Get a bunch of these automated cars hacked, and who's in charge of their "safer driving" with "no fatigue"? Think it can't happen?
Yes, I chose to compare the least competent automated system with the least experienced driver. then proposed that if the minimum standard is higher for computers, that's an improvement. It wouldn't be useful to compare the worst automated system with the average human. No weasels were injured in that explanation.
Humans certainly do learn! Even experienced drivers make stupid errors, though. And I'm not sure those mistakes are all novel ones. I suppose the automation manufacturer would "learn" during testing and real-world experience, program appropriate behavior, and then the behavior would be consistent across all cars equipped with that system, subject to updates. That's a scale of learning that's simply beyond human capacity.
Why not a 40 year old with over 20 years of driving experience? Presumably because the latter is a much better driver
Supposition much? How old do you think those wankers tailgating in BMWs are?
The biggest benefit to fully automated road transportation would be the higher throughput that would be achievable on the roads. This will be particularly noticeable in traffic jams or road works, where a lot of the slow down is due to driver uncertainty of what to do.
BTW there is a cure for tailgating... depending on the reason...
Reason (a) ... the most common.... get the f**k out of the overtaking lanes (middle and outside) when you are not f*****g overtaking anything. You may think you are doing the speed limit and I shouldn't be in a hurry but maybe I have someone in the car who needs to be in the hospital, don't run my life and I won't run yours!
Reason (b) .... next most common.... put your f******g foot on the accelerator and finish your overtaking manoeuvre in less than an hour - frankly you may want to saunter past the lorry but I want (or need) to get on!
Reason (c) ... the guy behind is an idiot in a BMW (or Merc or Audi... they have all spent more on their heap of German junk than it was ever worth but think they need to prove how huge their dick is to justify their stupidity in the showroom)... this is solved by a handful of ball bearings. You take a handful, put your hand out of the window, place the ball bearings on the roof, listen to the roll backwards and watch in the rear view mirror as the idiots windscreen is covered in nice little chips.
Have fun all :)
Reason (b) .... next most common.... put your f******g foot on the accelerator and finish your overtaking manoeuvre in less than an hour - frankly you may want to saunter past the lorry but I want (or need) to get on!
So when the speed limit is 70mph, and I overtake a train of lorries that are travelling at 60mph, you are the cockwomble behind me who feels that he's entitled to aggressively tailgate me to speed me up? And you are proud of this?
If you need to get to hospital quickly, call an ambulance. If you keep driving like that though, you'll need two, one for you and one for the poor sod you drove into.
".....a lot of the slow down is due to driver uncertainty...".... Really? I thought it was Rubber-Necking?
.... cant see an automated car slowing down to have a gawp at the tangled mess on the verge.....
... now excuse me while I read a novel, sip my beer and lean back whilst my car drives me home........
So it rather looks as if you think humans can learn from something, whether it be from mistakes or from simple practice. And yet you say they can't.
I have plenty of time behind the wheel and from that experience I gather that while people are capable of learning, they typically don't unless forced to do so. In the case of fatal automobile accidents, or fatal accidents of any sort, that presents a bit of a problem as the learning curve can be both steep and abrupt.
From a personal point of view, I don't have any sympathy for folks clinging on to their ability to drive when there are much, much safer alternatives available. Much in the same fashion that preventing people from smoking in most public places has improved the health of those who would otherwise be exposed to second-hand smoke, preventing people from inflicting now-unnecessary risk on everyone around them seems reasonable. To carry the analogy forward, perhaps to an absurd degree, we should next create roads just for human drivers isolated from the rest and let nature run its course.
AC, precisely the same problem as autonomous vehicles will have. How do you trust the sensors ? Pilots who fly by rules instead of evaluating all inputs tend to have more problems. In this case QANTAS pilots who had the same problems months before assessed the instruments as " Power normal cruise, angle of attack indicator normal, no engine anomalies therefore the airspeed must be wrong." and avoided trying to fix a non-existent problem. The less experienced AF aircrew in command seem to have believed one instrument, the airspeed and deep stalled the plane into the Atlantic. For these edge cases, increasing redundant information sources such as GPS would allow invalid inputs to be detected and cut out of control loops. Given how most airlines are entirely computer driven these days aviation automation still has outstanding safety record because unlike individual humans, the aviation industry has for decades had a learning and improvement culture that is still relatively new to consumer technology. The discussion then becomes whether it is computationally and sensor developmentally possible to improve vehicle situational evaluation. Given how the mobile phone industry has developed very light, cheap and accurate sensors I would hesitate to say it cannot be dome for vehicles. Whether enough development has been done is another story.
Sorry about polysyllabic words sounding like market droiding. Currently feeling crook.
"Flying is a piece of piss compared with driving. Nothing happens quickly, 1/2 km is considered a near miss so there is nothing to hit except the ground and it is pretty obvious where that is."
Quite a few accidents begin away from obstacles, and then "progress to the scene of the crash." I take your point, though, that driving in traffic has far less margin for error. A stranger in the next lane can lose attention for half a second and send 2 tons of Range Rover through your side door without warning.
I'm getting a bit elderly, although I'm not really there yet. I can still drive myself (and usually ride a motorcycle, so if I do fall too far below par and keep driving, Darwin will likely step in and take care of the problem). However, if I live long enough to get to where I truly can't drive safely, I would see a self-driving car as a Godsend, allowing me to still be independent and mobile (able get to the grocery store, etc.) without having to be a burden on others, asking for rides.
Why is this better than a driver doing such a job for you? It's an honest question.
I'm all for assistive aids, but I seriously doubt it is for this reason that AI driving is being pushed. The like of many are helped by Apple's assistive modes on the iPhone etc. But should they make something for those with need, the prices skyrocket (a screen reader/magnifier for the visually impaired etc, costs about as much as an iPhone that does even more and can do basic screen reading).
Don't think of yourself as a burden. :)
Well, mostly because saying it is and proving it is are two different things, and one needs to get it widely seen in order to change perception. AKA, how will the average Joe down at the coffee shop know that things are well sorted now if he doesn't read GIzmodo?
are suddenly concerned about traffic congestion and safety as well as the general well-being of the society. It warms my hart to see how technology giants care about us all.
Does anyone see anything wrong with this picture ? The only reason I see is those corporations have figured a way to extract money (and maybe something more) from you. It's a centralized vacuum pump in search of our wallets. It's all about control, folks!
However, there will be losers:
- luxury car manufacturers: why buy let's say a Porsche that drives itself ?
- insurance companies: why would I pay if it's not my car and I'm not the one driving it ? Maybe Google will do business with one major insurer in every country but the others will have to die.
- marketing boutiques: did anyone see an advertisement for taking a taxi ? All marketing for cars today is based on the pleasure of driving or feeling good behind the wheel. There will be no point in advertising one self-driving car over another since they will all be alike. You don't chose the model and maker of the taxi you called for a ride
- second car dealers: this one is obvious
- repair shops: we will not need that many, again maybe Google will do business with one or two large companies
- taxi drivers: file this one next to second car dealers
- health systems (especially in US): less people to fix, less revenue
Winners are:
- technology companies (big ones)
- those who are not able or willing to drive.
You're just parroting the same things that happened 100 years ago? What's happen to buggy manufacturers? Leather tanners for the buggy whips? Stables? Feed manufacturers and so on? Wouldn't you think they will just move on to other industries?
It's not parroting. Automated systems work well. They require on rails, or a large percentage automation (or asymmetrical).
We have none of that on the roads currently. In fact, it may reflect what happened 100 years ago. We went from having "roads" to having "roads and pavements" as the pedestrians were (literally) forced off the roads.
So it's not a problem as such... it's a requirement. The roads are required to change. Either "automation lanes" (or the opposite "manual lanes") to divide off the different types of traffic, or compulsory and extensive signage and driving method/style law changes.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
