German Auto Firms Face Roadblock In Testing Driverless Car Software

An anonymous reader writes As nations compete to build the first operational autonomous car, German auto-manufacturers fear that current domestic laws limit their efforts to test the appropriate software for self-driving vehicles on public roads. German carmakers are concerned that these roadblocks are allowing U.S. competitors, such as Google, to race ahead in their development of software designed to react effectively when placed in real-life traffic scenarios. Car software developers are particularly struggling to deal with the ethical challenges often raised on the road. For example when faced with the decision to crash into a pedestrian or another vehicle carrying a family, it would be a challenge for a self-driving car to follow the same moral reasoning a human would in the situation. 'Technologically we can do fully automated self-driving, but the ethical framework is missing,' said Volkswagen CEO Martin Winterkorn.

Comments

Cars!?

[Anonymous Coward]

Maybe the Germans should start focusing on pilotless planes.

Re:Cars!?

[prefec2]

First, this is outright cruel to say that. Second, this is only fear mongering of the car developers. They could test their cars in the US without any trouble and they have done the same in Germany. Yes they want to be allowed to put them on the road right now to show technology leadership, once, as they have been embarrassed by Japanese car manufacturers over the hybrid thing.

Re:Cars!?

[cayenne8]

Maybe the Germans should start focusing on pilotless planes.

Oh my God....my kingdom for Mod points for this one...

ROTFLMAO....

Re:Cars!?

[recharged95]

That's a deep answer. This is the funniest article of the day for me as a flying robot/drone developer.

While the germans are flying drones all over the place, selling them and have a regulatory framework, they are complaining they cannot build/sell autonomous cars... and calling out other countries, mainly the US as beating them in that game.

But in the US, it's the completely reverse... or bizarro situation. While Google and Uber are building autonomous cars, and getting them approved for use, drones on the other hand are DOA with no framework insight... and drone companies are complaining countries like Germany and France (and Australia) are beating them in that game.

Oh, "the irony of the rant".

Re:Cars!?

[Anonymous Coward]

For those who don't know, auto pilots can already fly every aspect of a flight. Auto pilot is so good they had to add randomness to it as planes kept touching down in the exact sames spots on the runway, quickly wearing out those areas instead of spreading the wear around. Human pilots handle emergencies, bad sensors, and make every feel better.

Re:Cars!?

[Anonymous Coward]

I wish you get modded to hell for this.

Re:Cars!?

[Anonymous Coward]

Bullshit
There is no serious drone company in Germany.

Re:Cars!?

[Anonymous Coward]

Companies are always complaining and comparing the incompetence of the local government to the always greener grass in other countries.

Not concerned

[Anonymous Coward]

Pretty sure autonomous vehicles will be banned in all 50 states.

Re:Not concerned

[afidel]

I think it's more likely we'll ban human drivers. Just this morning I counted over 16 silver/grey/blue-grey vehicles driving in pouring rain and light fog without headlights on. On average a computer driver today is probably better than a human, and they'll just get better as time moves on whereas human improvements are a bit slower to happen.

Re:Not concerned

[TWX]

The trucking industry would absolutely love to do away with hundreds of thousands of long-haul drivers. The mass-transit industry, which is often contracted by the municipality to private companies, would also love to do away with bus drivers and other high occupancy drivers where they could be deemed not necessary. On top of that, removing the payroll for drivers could allow bus companies to start employing private security on bus routes where assault or vandalism is a problem without increasing their payrolls to do so.

Even low occupancy transit like taxis will do away with drivers- it will remove the human element as a risk to the passenger and will mean that the cab companies make more money as they're not simply renting cabs to drivers for a flat rate, they're collecting all of the revenue for the cab's use, and they only have to operate as many cabs as they have service demands for at any given moment, so there's less unnecessary wear and tear on the cars as drivers aren't speculatively taking cabs out.

Sure, there will be plenty of human drivers out there, but there's going to be a whole lot of automation because it will simply be much more cost-effective in many circumstances.

Re:Not concerned

[Coren22]

That will become a union thing (in both cases). The trains and trucks will still be required to have a warm butt in the seat, the butt will just be totally ignoring the windshield though instead of just partially ignoring.

Re:Not concerned

[ShanghaiBill]

The trucking industry would absolutely love to do away with hundreds of thousands of long-haul drivers.

At least in America, the drivers are the trucking industry. When you see an 18 wheeler on the freeway, the chances are very high that the truck is owned by the guy driving it.

Re:Not concerned

[Anonymous Coward]

Which also could benefit consumers as prices could be lowered without the need for driver salaries. Not that this is likely, but it is something consumers could push for since there would be an obvious reason for price reduction.

Re:Not concerned

[TWX]

Who owns the trailers and the merchandise? It's usually not the owner of the tractor.

Who pays the trucker? The owner of the merchandise or the trailer.

Don't forget, lots and lots of large retailers maintain their own over-the-road fleet. Sears/Kmart, Walmart, Target, Costco, Kroger, Safeway, Autozone, and that's only a drop in the bucket. They could all retrofit to an automated tractor, or at least where a pilot car or truck escorts a caravan of autonomous trucks following behind.

Re:Not concerned

[Anonymous Coward]

Pretty sure autonomous vehicles will be banned in all 50 states.

As evidenced by the zero states that have made even a token effort to ban them?

Re:Not concerned

[Anonymous Coward]

Replacing trucks with trains would improve things.

Re:Not concerned

[Anonymous Coward]

The trucking industry would absolutely love to do away with hundreds of thousands of long-haul drivers.

The trucking industry would rather get rid of short-haul drivers first. It's far more labor-intensive to get a 10 lb package to your doorstep than to haul 20 tons across the country.

Re:Not concerned

[TWX]

Trains are great for large volumes to dedicated yards where the terrain allows the train to pass. Unfortunately there are lots of towns and cities where railroad access is poor or nonexistent. There are also deliveries of perishable goods from regional distribution centers to retail locations, like grocery stores, where they cannot afford the time to load, move through the 'backbone' of the rail system, to then unload at the next railyard, to then be loaded on to a truck for final distribution.

I can use my own commute as an analog; I drive around ten miles to work, going a half-mile east to the freeway, going a mile and a half north on that freeway to another one, about five miles east, then about three miles north. It takes about fifteen minutes if I hit all of the lights red. If I were to take mass transit I would have to wait for a bus to go half a mile WEST, then go north about six miles by bus, then take another bus east that goes the six or so miles but makes a detour to a senior center and sits for fifteen minutes, then walk a quarter mile up to the business. it would take me well over an hour even if I hit all of the connections perfectly. It makes more sense to drive.

Re:Not concerned

[TWX]

Someone still has to unload the package and bring it to the doorstep. Packages are not of uniform size. It's much more likely that long-haul, with depots at both ends where robots and humans can load the trucks efficiently and completely unload them at the destination, will allow for an autonomous navigation due to a lack of needing humans in between. A route with numerous stops will be much harder to automate without adding a whole bunch of equipment to the trucks.

Re:Not concerned

[Anonymous Coward]

You might actually be surprised at who does or does not operate large fleets of their own trucks. I worked for Costco, for example, and every single one of our deliveries came on a Swift truck. I suspect that it's actually pretty common for quite a few of those large companies you mentioned NOT to run their own fleet.

Re:Not concerned

[EvilSS]

Don't worry, they can all get new jobs servicing the self-driving vehicles that replaced them. That's how it works, or so Slashdot tells me.

Re:Not concerned

[Anonymous Coward]

The Trucking industry aren't the people who want things shipped. The Trucking industry are the people who ship the things, e.g. the trucker. The guy who owns the truck is the trucker, not the company who owns the trailer. Any idiot can own a trailer and load it up.

Re:Not concerned

[gmhowell]

Most large companies outsource their transport to JB Hunt, Schneider, etc. Sure, the big letters say 'WalMart', but in smaller, DOT minimum sized font, it often has another name.

Re:Not concerned

[drinkypoo]

Most large companies outsource their transport to JB Hunt, Schneider, etc. Sure, the big letters say 'WalMart', but in smaller, DOT minimum sized font, it often has another name.

So what you're saying is that there are actually fewer transport companies than there appear to be? Because that's an argument in favor of self-driving trucks, once again.

Re:Not concerned

[TWX]

I think that only a fraction of the employed drivers could turn to providing service or operating pilot trucks if there's an interest in using automated convoys with a real driver or two to babysit. It's one thing to change a wheel when a tire blows (even that usually requires existing service companies to find the truck to assist) but dealing with other problems requires trained mechanics. Driving, by contrast, is much easier than that.

I expect that some human drivers will remain for odd jobs, like weird oversized loads that are poorly balanced, or routes with poor or unimproved roads, or the logging industry where the trucks are operating off-road, but there's going to be a whole lot of people looking for work.

Re:Not concerned

[ShanghaiBill]

Because that's an argument in favor of self-driving trucks, once again.

Nobody is arguing that about whether SDTs are coming, but about who will drive the change. The "trucking industry" is unlikely to be an agent of change. They are entrenched incumbents who will fight, lobby, and bribe to stop automation. Progress is more likely to be driven by customers such as WalMart, or entirely new transport companies. They will be lobbying and bribing in the opposite direction.

Re:Not concerned

[drinkypoo]

The "trucking industry" is unlikely to be an agent of change. They are entrenched incumbents who will fight, lobby, and bribe to stop automation.

Again, why would they do that? The only members of the "trucking industry" who stand to lose if trucks go automated are truckers themselves, but they can't possibly out-lobby trucking companies.

Progress is more likely to be driven by customers such as WalMart, or entirely new transport companies.

No. Wal-Mart will just contract with whoever can moves the trucks most cheaply. But they're not going to do the leg-work themselves. They'll just contract whoever has the self-driving trucks, after they do the lobbying.

Wal-Mart doesn't give a shit how cheap trucking is, only that they get it at the lowest possible cost, because they don't care what their prices are, only that they are lower than the competition.

Re:Not concerned

[ShanghaiBill]

Again, why would they do that? The only members of the "trucking industry" who stand to lose if trucks go automated are truckers themselves, but they can't possibly out-lobby trucking companies.

Again, the drivers are the trucking companies.

Re:Not concerned

[Anonymous Coward]

The guy who owns and drives the truck is not going to be able to compete with the company that owns self driving trucks.

Re:Not concerned

[TWX]

JB Hunt, as an example, is who is contracted to move goods. They have close to 50,000 trailers. It appears that they own around 12,000 trucks. Even if they own zero trucks now and are entirely dependent on independent owner/operator contractors and those contractors' trucks, they could start purchasing their own self-driving trucks and no longer contract-out for drivers and their trucks.

Large companies like doing business with other large companies. If JB Hunt, Swift, Allied, or any of a slew of large trucking companies stop using human drivers then the big players like Walmart, Costco, Target, all of the department stores, will simply continue to use JB Hunt et al. because they're not going to contact thousands upon thousands of independent owner-operators. They care about selling merchandise. If they're not already running their own fleet directly then they'll continue to sub that out to whoever can.

Re:Not concerned

[mjwx]

The trucking industry would absolutely love to do away with hundreds of thousands of long-haul drivers.

At least in America, the drivers are the trucking industry. When you see an 18 wheeler on the freeway, the chances are very high that the truck is owned by the guy driving it.

Its the same in Australia.

Truckers are a very powerful union and a lot of them are owner/drivers. This is changing of course, but it's going to be a slow and laborious process. However it will be a very long time before we have fully autonomous trucks.

Re:Not concerned

[Anonymous Coward]

ould allow bus companies to start employing private security on bus routes where assault or vandalism is a problem without increasing their payrolls to do so.

They don't have to increase payrolls for security since they just buy the security services from a contractor, like most businesses do.

Re:Not concerned

[sl149q]

Self driving vans will be deployed by UPS / Fed Ex / et al simply so that they can have the driver become a full time package sorter and deliverer. On some routes in a busy downtown area there may even be multiple people in the van getting dropped off and picked up by the self driving van which then does not need to park.

You could also see people moving from an empty truck to a newly arrived full one. The empty one heading back to the depot to fill up. The full one having driven itself out from the depot.

People will still be involved. They just won't be driving the vans around. Using them optimally will be a competitive advantage.

Re:Not concerned

[Anonymous Coward]

Is his name 'Rubber Duck'?

Re:Not concerned

[gmhowell]

I should actually correct myself slightly: Wal-Mart (and others) have some in house drivers and some outsourced.

BTW, in discussions of the transport industry, don't get distracted/lied to by the companies. Some drivers think they are owner operators, when in practice, they aren't. They will lease/buy a truck from (as an example, all of the bigs do this) Schneider. As part of the lease terms, they can only accept loads from Schneider. It should be obvious that the 'owner' is an employee who has assumed much of the risk that the company would usually take on.

ShanghaiBill has a decent reply, but he misses a point: if the automated truck is cheaper, the big companies will drive that change in a heartbeat. The trick is that someone has to be convinced that they will be cheaper. They are unlikely to automatically accept that an automated truck is safer, faster, etc. One area where they are likely to be impressed is the possibility of 24 hour operations, rather than the 10 hour per day (rough) limits of human operated trucks. In addition to (possibly) being cheaper, this will allow faster shipments for more mundane goods (there are already plenty of ways to have fast shipping, but it is cost prohibitive to do for everything) which would offer them a competitive advantage. I suspect this last point will be the thin edge of the wedge.

Re:Not concerned

[drinkypoo]

Again, the drivers are the trucking companies.

A romantic notion, but incorrect. Private fleets account for 80% of trucks and over 50% of OTR tonnage shipped.

Re:Not concerned

[TWX]

That lease and exclusivity scenario completely destroys ShanghaiBill's argument about owner-operators if they are indeed at the mercy of the company to whom they're contracted.

I expect lower operating costs could come from simply not having to operate as many trucks. If man-operated trucks usually only operate 5/12 of the day, it's conceivable that autonomous trucks could operate much closer to the full day, less maintenance, refuelling, and load/unload or hookup/unhook, and those latter tasks might count toward the ten hours that a human operator is allowed to work. Depending on routes a trucking company might only have to operate half as many tractors to pull the same number of trailers around, so capital costs and insurance costs could go down rather dramatically. Even if there are scenarios where trucks are organized into convoys and a human-operated pilot car or lead truck is necessary, it still might be possible for 20 loads to be operated by a single driver, or if 24-hour service is wanted, three drivers working in shifts on the convoy, trading off the pilot position.

Labor costs are usually the biggest expense to an organization. They're always looking to reduce labor costs, so I fully expect this to come.

Re:Not concerned

[gmhowell]

There are some owner operators as described, but most are beholden.

The problem with swapping drivers has to do with hours of service requirements. A bit too much detail to get into right now, unfortunately.

Re:Not concerned

[TWX]

I expect that since Australia is so much more rural than even the western half of the continental United States, there will be more need for humans to be involved with trucking simply to handle situations that crop up outside of the computer's ability to handle.

How closely can road-trains operate to each other? I could see a lead truck with a human in it even if it's still driving mostly autonomously, a bunch of fully autonomous trucks, and a tailgunner with a human too, so that if there's a problem far down the line in such a convoy someone would notice it even if the computers didn't. Plus, if those drivers are responsible for tires and other on-the-road maintenance, having more than one person might be handy.

Re:Not concerned

[TWX]

That sounds pretty good actually.

Apparently back in the day, telecom workers for the phone company didn't have their own service vehicles. One van or small bus would drop workers off at their sites so they could splice cables, which takes awhile, and would then collect them after they finished, to move on to the next area. After build-out has ended it has become more necessary for workers to have their own service vehicles as they don't need to drop five workers off along a ten mile stretch to each work for two hours.

For doing dense urban package delivery this does make sense. It might still make sense for one of the occupants to be licensed to drive the truck though, so that if they have to override the computer to park it they're legally allowed to.

Re:Not concerned

[TWX]

It's still money though. You either pay your own staff or you pay someone else to pay their staff.

Black and White

[Murdoch5]

Just weight the situation, if the car you're in is going 40 km/h and you know that you can slow to 20 km/h then hit the car in front of you because the pedestrian won't survive. Also you'll have access the license plate, so you can easily look up information on the family in the car and get the stats of everyone in the car. More then enough data will exist at the time at the accident to judge it properly, what is the issue? You'll have more information then any normal driver would. It really is just black and white.

Re:Black and White

[Anonymous Coward]

Black and white?

What is the pedestrian is jaywalking. Hitting a car means your insurance is ruined, two vehicles may be totalled, you can end up with a dangerous driving charge, and overall may ruin your driving career. Hitting the pedestrian means some very minor damage to your car is repaired at their expense, they have several broken bones which will heal, and learn why jaywalking is a crime.

It's a very grey situation and the choice depends on many factors.

Re:Black and White

[Murdoch5]

Well except that in Ontario Canada, you can't hit a jwalker as you'd still be totally at fault. Pedestrians have total and absolute right of way, you'd have almost a 0% chance of fighting it in court.

Re:Black and White

[roman_mir]

it doesn't matter who is at fault, you hit a pedestrian, he or she is likely to die, so try not to hit them, however it's not all that cut and dry in Ontario either, people sometimes get killed on the road and drivers are not always charged with anything.

Re:Black and White

[Anonymous Coward]

What is the pedestrian is jaywalking

Is that where they just jump out right in front of you, sort of like the telephone pole you hit the other day?

I won't use an autonomous car until it can determine that there is something not-a-car in view in less time than a human takes to do so. Ideally, it would identify the moving meatsack the instant it steps into the road and already begin driving defensively (eg slowing down and pulling around the car everyone insists it must slam into at 100mph) before a human would have even reacted.

Re:Black and White

[khasim]

It's even easier than that.

Do YOU want to be the person dragged into court because YOU wrote the program that INTENTIONALLY HIT AND KILLED someone?

No? Then write the code to be 100% neutral. The code will ONLY attempt to stop the vehicle as fast as possible.

If pedestrians are within X meters of the car then the car should slow to Y. If they get closer then the car should stop.

But the code should NEVER have the option "hit object X".

Re:Black and White

[Reaper9889]

So a pedestrian is walking on the sidewalk 10m from your car. Do you slow down? If not, how will you avoid hitting him if he suddenly decides to sprint and jump infront of your car? If you do slow down, people will never use the autodrive feature.

You could write it under the assumption that everybody else was trying to make you fail (basically it would become a zero-sum game) but it is fairly clear that it is far too restrictive in any real sense.

Re:Black and White

[Murdoch5]

Well being that I develop embedded level software that could deal with this kind of problem, I can actually summarize exactly how I start:

Lets just assume the object we're concerned about is in front of you!

Step 1: Turn on all cameras and sensors
Step 2: Cut all engine power
Step 3: Using the license plate of the car in front of you determine what would happen if you hit it.
Step 4: Compute what would happen if you hit the person
Step 5: What ever has less damage execute
Step 6: Dump massive logs to solid state storage
Step 7: Call police, send them the files automatically
Step 8: Run diagnostics to determine driver error or manufacturing issue
Step 9: Done.

Not that hard to do ....

Re:Black and White

[khasim]

If not, how will you avoid hitting him if he suddenly decides to sprint and jump infront of your car?

That would be "suicide".

And the sensor logs of the car should be able to show that it was suicide.

But more to the point, how would that situation be any different in a faster-reacting-autonomous-car than in a human-controlled-car?

Or are you postulating a world where there are no cars because someone might try to commit suicide by jumping in front of one?

Re:Black and White

[khasim]

Which of those steps covers engaging the braking system?

And how does the license plate "determine what would happen if you hit it"?

Why weren't the cameras and sensors on already if the car was operating autonomously?

Re:Black and White

[Murdoch5]

Well because the mass amount of data that would be grabbed in the event of an accident would far overshadow a reasonable amount of capture memory during normal driving, which would utilize a lesser set of sensors and maybe lower grade video, which didn't have to factor into the explanation for the accident.

Step 2 would include choices such as hit the breaks if it would work. I just used summary steps to make it easy to understand.

Your plates store information about your car, hence you know from looking the number up, everything to know about the car via reference lookup. Wireless is easily fast enough to deal with the latency for it so no problem there, infact you could also get information about the driver, so really you do have a good deal of information at your car to decide what to do.

If the software fails for some reason, a good hardware design will step in to save it so again no problem :-)

Re:Black and White

[khasim]

Well because the mass amount of data that would be grabbed in the event of an accident would far overshadow a reasonable amount of capture memory during normal driving, which would utilize a lesser set of sensors and maybe lower grade video, which didn't have to factor into the explanation for the accident.

256GB of flash is just over $100 right now. Storage is not a problem. Even AIRCRAFT do not have a problem with storage and they have a LOT more data to store.

Step 2 would include choices such as hit the breaks if it would work. I just used summary steps to make it easy to understand.

Taking power from the engine is NOT the same a braking.

Taking your foot off the gas is NOT the same as stepping on the brake.

Seriously. Try it on a hill. You might end up going FASTER at the bottom of the hill than at the top.

Your plates store information about your car, hence you know from looking the number up, everything to know about the car via reference lookup.

Make/model/year/VIN/owner/owner's address. And maybe whether it passed inspection or not.

How will knowing the VIN tell you anything about hitting it?

Or the owner's address?

Or the owner's name?

Or any of the other information?

And what happens when the site you're trying to use to look up that useless information is slow?

Re:Black and White

[Anonymous Coward]

...You're overcomplicating it.

ALWAYS choose to hit the car, preferably at an angle that maximizes the moment of impact. That simple rule will kill less people.

Re:Black and White

[Anonymous Coward]

I'd say you're wrong.

Sure, you never want to hit either a car or a pedestrian if you can avoid it. If, however, a collision is inevitable -- it's impossible to avoid -- then you can still minimize damage a great deal by properly managing the impact.

For example: you're travelling on a two lane road with a median to the left of the left lane and a wall that leaves no break-down lane on the right. You brake, but you hit a patch of black ice and start to skid. If you just start braking then you're going to rear-end the car in front of you. If, however, you start braking and aim at the middle of the two cars, hitting both of them and basically wedging yourself into the middle -- you're going to total your car, but you're also going to maximize the moment of impact, minimizing the g-forces that all parties in all cars are exposed to, and minimizing the chances of bodily harm to all involved.

This happened to me when I first started driving. I hit the cars -- which were at a dead stop -- doing about 30. No injuries all around.

Re:Black and White

[Anonymous Coward]

As an Ontarian, I know this is bullshit. Pedestrians do not have right of way unless they are following the rules. There are many cases in Ontario where drivers have killed Jaywalkers and have not been charged.

The rule in Ontario is actually that you must take due care to avoid a collision with a pedestrian. However, due care does not involve running into other things if those things also have people in them. For example, if your only choices are to run into a jaywalker or drive through a restaurant, you will not be charged with a crime if you kill the jaywalker. Also important is if you avoid the jaywalker and damage property in doing so, the jaywalker will be paying for the damage, should they be caught. And if you hit the jaywalker, they will be paying to fix your car.

To prove you're bullshitting, here it is, straight from a police department in Ontario:

http://www.gsps.ca/en/specializedservices/pedsafety.asp

"Watch out for jaywalkers. Although drivers have the right of way over jaywalkers, in the event of a collision no one wins."

Re:Black and White

[Anonymous Coward]

>Is that where they just jump out right in front of you

Pretty much. I drive at the speed limit past the local university. At 60 km/h, through intersections where the foolish students ignore red lights and walk in front of cars. We have a saying that the students learn to walk responsibly after the first one of the year is killed in an accident (and it does happen, every single year).

>sort of like the telephone pole you hit the other day?

Wow, in your country a telephone pole will walk into the path of a quickly moving object? This is on some other planet, correct?

Re:Black and White

[sl149q]

Its postulated that the teenagers of tomorrow may have a new fun game that involves playing chicken with autonomous cars on the local freeway. Just walk across and watch all the cars veer away and avoid hitting you.

Avoiding them is no different from avoiding other moving targets such as dogs, deer, moose, etc. Anything that might damage the vehicle or occupants needs to be avoided. With the message (moving obstacle on road) being broadcast to the rear for following cars and the front for approaching (in the opposite lane) cars to also avoid.

In theory at least if you are visible at the side of the road then all of the cars on that stretch of road will be aware of your position and movement vector and the probability (updated in real time based on your observed movements) that you might enter the road way. The only way to get onto the road and "surprise" a robotic car will be to hide behind something and leap out at high speed. Or possibly off something.

Re: Black and White

[Anonymous Coward]

indeed. Just in case anyone thinks this is far fetched, we already live in a world with human drivers forced into errors by other humans dropping bricks off a bridge.

Re: Black and White

[Anonymous Coward]

How do know you slowing down is the right thing to do? It often isn't. Lots of motorway/freeway accidents would be avoided if people changed lane instead of hitting the brakes in a panic. The original principal behind ABS is to allow the layman better steering under braking.

Biggest issue is still liability

[gstoddart]

So, disregarding how the self-driving car decided who it is best to kill in any given situation, for me the biggest problem with self-driving cars is legal liability.

If Google wants to sell autonomous cars, Google should be liable for anything the damned thing does.

And none of this cop out where if the computer doesn't know what to do it just hands back to the human -- because that's pretty much guaranteed to fail since the human won't be able to make the context switch in time (if at all).

As far as I'm concerned, the autonomous car has to be 100% hands off by the user at all times, and the company who makes the damned thing is 100% responsible for what it does.

Why the hell would someone have to pay for insurance for something they don't have control of what it does?

Re:Biggest issue is still liability

[Coren22]

So what happens when the brake assist in a modern car causes an older car to rear end a car on the highway? Who is liable as it was technology that caused the accident, not either drive.

These things are already dealt with in modern countries, but lets pretend that all the years of driving related liability rulings never happened.

Re:Biggest issue is still liability

[afidel]

Why the hell would someone have to pay for insurance for something they don't have control of what it does?

Says every parent of a teenager since cars became widespread.

Re:Biggest issue is still liability

[HornWumpus]

Same thing that happens when a modern car with brake assist rear ends an old car with better brakes and traction.

If your car has shitty brakes you leave extra room. Good drivers realize that 'shitty brakes' is always relative.

If you're driving a crazy high performance car you moderate your brake use to avoid being rear ended.

Re:Biggest issue is still liability

[gstoddart]

Except they're not the same.

If you have a human driving, you usually know who to blame.

If you have a computer driving, the people who made the computer sure as hell aren't going to take liability.

But you can bet your ass some sleazy lawyer will put it into the EULA that by driving in a Google car you assume all liability.

If they're going to make autonomous cars, they pretty much need to be 100% autonomous, with the humans effectively in the back seat with no controls.

At present, there simply ARE no liability rulings about self-driving cars.

Re:Biggest issue is still liability

[TWX]

There's always a degree of responsibility and liability if one owns something. I own a ladder. If someone uses that ladder at my house with my permission and the ladder breaks and injures them, even if it has not been abused, my homeowners' insurance policy is probably going to end up having to pick up the tab.

I fully expect that insurance for completely autonomous cars will be less expensive, once self-driving cars are proven. To prove them, I expect large fleets sponsored by the manufacturer or systems integrator will drive many thousands of hours per-car to establish a baseline, similarly to how an MTBF is established for devices, and that rate of collision or other liability-causing event will factor into the insurance companies' rates for those cars.

For autonomous cars that are capable of being human-driven, my guess is that insurance will be based on a percentage of on-road driving done by the computer versus on-road driving done by a human, and then the class or category that human falls into would apply for that ratio.

Re:Biggest issue is still liability

[suutar]

So, falling back to first principles... the following car should be prepared for the lead car to do pretty much anything, including drop a bumper (which will come to a stop far faster than either car can brake). If you're not leaving enough room to avoid hitting a fallen bumper, you're too close. Follower at fault, next case.

Re:Biggest issue is still liability

[Coren22]

A dropped bumper will tend to slide along the ground and likely will go off the side of the road before the approaching car gets anywhere near it (coefficient of friction of steel or plastic vs rubber will show you that), however, when a tire falls off the lead car and the disc brakes are instead digging into the road surface, that will stop rapidly.

However, that was just an example. There are others for current "autonomous" features in cars. How about the speed adjusting cruise control (when it malfunctions and the user doesn't realize in time to brake the vehicle)? Or even 4 wheel drive (honestly officer, I had no idea my truck would accelerate that quickly that I would run over that poor little old lady). ABS (it actually reduces your braking in snow, and is less than useless on ice). Traction control (can cause you to not be able to move AT ALL in snow and ice conditions...this could be a bad situation).

Re:Biggest issue is still liability

[ShanghaiBill]

for me the biggest problem with self-driving cars is legal liability.

Why is this a problem? Several states already allow self driving cars on the road (although with a driver in the seat for now). The liability issue is already resolved. The party responsible is the insurance company. Duh.

The only thing that changes with SDCs, is that the insurance will likely be much cheaper. Not just because accidents will go down, but also because the camera, gps, and sensor data will make it very clear what happened, eliminating disputes over the facts, so legal costs will be much lower as well.

Re:Biggest issue is still liability

[Anonymous Coward]

I would imagine that some of the accident would still be genuine accidents - sensors failing, unintentional interference, and plain situations where there is nothing possible to do because it is impossible to stop from 100 kmh to 0 kmh in 5 meters. Also the image recognition algorithms may fail in some peculiar situations, which would cause 99.999% of humans to fail as well. Who is liable then? Entropy? God for making such un-intelligently designed universe (I suspect unintelligent design, rather than intelligent design)?

Re:Biggest issue is still liability

[gstoddart]

I fully expect that insurance for completely autonomous cars will be less expensive, once self-driving cars are proven.

Again, why would I pay liability insurance to cover the actions taken by a computer?

The only viable business model for fully autonomous cars I can see is essentially as taxis.

The notion that we're all going to trade in our cars and let the computer do all the driving is laughable -- too many people like driving, and there's decades worth of cars out there. The notion that we'd buy a self driving car and then pay to insure it is silly.

As with most things said by futurists, I find myself wondering if this "inevitable future" is anywhere near as inevitable as the people selling it to us want us to believe, or if it's just more marketing of stuff someone wants to sell us.

My take on this is the people making these things claim how awesome and revolutionary they'll be. But I bet the vast majority of people simply don't care.

Re:Biggest issue is still liability

[K. S. Kyosuke]

If your car has shitty brakes you leave extra room. Good drivers realize that 'shitty brakes' is always relative.

Sounds to me like the solution to the problem in question - a computer could quickly periodically recompute the envelope of possible scenarios and never drive in the phase space into points from which it can't recover without hitting someone or something.

Re:Biggest issue is still liability

[K. S. Kyosuke]

If you have a human driving, you usually know who to blame.

Which, to me, is a horrible way of looking at things. If that were the only criterion, we could easily end up with ten times more car deaths simply because we're more comfortable with putting blame at people, even at the expense of lives.

Re:Biggest issue is still liability

[itzly]

The point is moot. Either the owner pays the insurance, or the owner pays Google, and then Google will pay for the insurance.

As long as insurance companies are willing to provide the insurance, the finer liability issue isn't important.

Re:Biggest issue is still liability

[itzly]

Again, why would I pay liability insurance to cover the actions taken by a computer?

Because that's the most pragmatic solution. If you don't like it, don't get a self driving car (and probably pay even more insurance).

Re:Biggest issue is still liability

[HornWumpus]

That computer would stop on the on ramp of SF bay area highways and refuse to move.

Re:Biggest issue is still liability

[gstoddart]

Welcome to a world with lawyers and liability laws.Someone is always to blame.

And, as I said, you can bet your ass Google et al are going to try to make sure it's you and not them.

Re:Biggest issue is still liability

[bws111]

Huh? It certainly is the fault of the driver in the older car.

Re:Biggest issue is still liability

[khasim]

To prove them, I expect large fleets sponsored by the manufacturer or systems integrator will drive many thousands of hours per-car to establish a baseline, similarly to how an MTBF is established for devices, and that rate of collision or other liability-causing event will factor into the insurance companies' rates for those cars.

I think it will be even easier.

The autonomous cars will be packed with sensors that record EVERYTHING.

If there is an accident then the insurance companies will know which car has a 100% complete record of the incident that SHOULD exonerate it. Such as staying below the speed limit. Keeping a recommended distance from the car in front of it. Staying in the center of its lane. And exact information on how hard the brakes were applied and when and how that affected traction prior to the collision.

In theory, the insurance company for the autonomous car should win ever time (except in cases of software/hardware failure).

Re:Biggest issue is still liability

[bws111]

Every single one of your examples is just bad driving, and have nothing to do with autonomous features in cars.

Cruise control malfunctioning is no different than the car in front slowing. It is your job to notice and react. If you 'don't have time' then you are tailgating.

4 wheel drive? What does that have to do with acceleration?

ABS - yes, that is what is does. If you are driving in snow, leave more room. Not that complicated.

Traction control - yes, that is why it has an OFF switch

Re:Biggest issue is still liability

[Coren22]

Did you miss the first two words on the Cruise control line? Speed Adjusting Cruise Control, it is the new hot thing, it adjust the speed of the vehicle automatically to the vehicle in front. If it were to malfunction you may not know in time to react.

4 wheel drive ONLY effects acceleration, what else would it have control over?

You are missing the point, all of these systems (besides 4WD...) are being integrated into an auto driving car, so we have to look at the failure modes to understand where the liability is (the article this was all replying to).

Re:Biggest issue is still liability

[bws111]

You seem highly confused as to what insurance is. Here is a clue: the 'party responsible' is NEVER the insurance company. The 'party responsible' is YOU, the insurance company is just providing the cash for you.

Its cute that you think cameras, gps, and sensor data will make it very clear what happened or eliminate disputes.

Re:Biggest issue is still liability

[Coren22]

Not always. If I slam on my breaks (ABS, electronic brake distribution, traction control, etc) and someone rear ends me, I could be at fault for overbraking even if it was needed to prevent me hitting someone/something.

Re:Biggest issue is still liability

[bws111]

It does not matter if the speed control is speed adjusting, it is still YOUR responsibility to maintain safe distance.

4 wheel drive does not affect acceleration, it affects traction. The only way to blame 4WD for unintended acceleration is if you were planning on spinning your wheels, which again is just shitty driving.

And, no, I am not missing the point, you are. The 'failure modes' you listed, whether or not some technology was involved, are failures of the DRIVER, and that is where the liability resides. That does NOT tell us who will be liable when there is no driver.

Re:Biggest issue is still liability

[bws111]

Where? Everywhere I have been the driver in back is always at fault in rear-end collisions.

Re:Biggest issue is still liability

[Anonymous Coward]

I'm almost positive that it's not nearly that simple as, if the bumper falls off of your car, short of some spectacular, catastrophic component failure you, as the driver, probably knew or should reasonably have know that there was something wrong with your vehicle which well may cause you, as the driver of the fault vehicle, to be liable for any damage caused by your improperly maintained vehicle.

Here's an even more fun one: bumper falls off of a vehicle ahead of you but in another lane. You are following at a more than safe distance from the car directly in front of you; however, the car with the faulty bumper was overtaking you on your left. There is no distance keeping requirement for cars not in your lane and, especially, not when the other vehicle is legally passing you on your left. The bumper fails, falls off of their vehicle into your lane causing an unavoidable collision. You are not at fault.

A more tame example which actually happened to me: I was behind a truck. The truck had a rack of lights on top. One of the lights supports had rusted through and completely disengaged from the top of their vehicle. This light hit my windshield severely damaging it. Was I at fault. No, I was not. Why? Because ultimately the other driver failed to properly maintain their vehicle so the failure of the component and the damage caused was their fault.

Re:Biggest issue is still liability

[deadweight]

FYI ABS brakes were available in Europe for years before the USA *for just that reason*

Re:Biggest issue is still liability

[trekdvorak]

It is the fault of the car that CAUSES the accident. The fact that an older car was involved is irrelevant as this is just another variable in the driving environment that the driver of the newer car seemingly failed to account for. The build date of the car should not imply fault unless you want to repurchase your vehicle on a yearly basis to avoid being liable. Doesn't sound like a whole lot of fun to me and probably isn't sustainable.

Re:Biggest issue is still liability

[trekdvorak]

Go ahead and try that logic and let us know what happens. It's been tried before and is often first looked at as insurance fraud.

Re:Biggest issue is still liability

[ShanghaiBill]

The 'party responsible' is YOU, the insurance company is just providing the cash for you.

As long as the insurance company is paying, why should I care who is "responsible"?

Its cute that you think cameras, gps, and sensor data will make it very clear what happened or eliminate disputes.

In many countries, insurance companies offer discounts for anyone using a dash cam. Why? Because cameras reduce disputes, thus lowering legal costs.

Re: Biggest issue is still liability

[Anonymous Coward]

There was a woman charged with manslaughter for stopping for a rabbit and causing many human deaths instead.

Re:Biggest issue is still liability

[bws111]

Insurance fraud is not an accident, it is a crime. Big difference.

Also, I should have mentioned rear-end accidents caused by the lead vehicle 'stopping short'. That is always the fault of the following vehicle for not leaving enough space. There are cases where the lead car may be at fault, such as merging in too close to following cars, but that is different from just stopping short, which is what was originally discussed (brake assist causes car to stop quicker than following car without assist).

Also, there is a big difference between fault for an accident and compensation for an accident. In the insurance fraud case, the following vehicle is still 'at fault', but the leading vehicle may be denied compensation if fraud is suspected.

I didn't find any references to 'overbraking' being against the law, while following too close clearly is.

Re:Biggest issue is still liability

[TWX]

Cars are very complex machines that can have loads of things go amiss with them without rendering them undrivable, and can have loads of other things go wrong while in operation. You're correct that it's safe to assume everything will be recorded, but I expect equipment failures will plague first-generation autonomous cars once they're old and the tolerances have loosened up. Steering, tires, brakes, suspension alignment, all things that will lead the computer astray as it's attempting to self-drive.

Re:Biggest issue is still liability

[Anonymous Coward]

Just going to point out since it originally dealt with liability issues we remain in the exact same place as it all started. Autonomous cars need to be reliable to the point where the passenger isn't liable or else there's no point to them.

Re:Biggest issue is still liability

[suutar]

For cases where the only difference is who was making the decisions, I'd say liability should be with the manufacturer. Other cases are similar to existing with-driver cases: parts failure (manufacturer), skipped maintenance (owner), poorly performed maintenance (shop that did the work).

In the end, like with FAA investigations, everything boils down to pilot error and equipment failure, and in many cases "not dealing with equipment failure properly" is considered pilot error. The only question is who's the pilot.

Re:Biggest issue is still liability

[Anonymous Coward]

Because it's never the insurance company that's paying? What, do you think insurance companies are run by really rich people who just give their money away to people because they need it? No, you give money to really rich people who horde it, shave some off for themselves, and in the event of you needing some of that money back that you've given them, work tirelessly to figure out how to give you as little as possible.

Seriously, have you never worked with an insurance company before?

Re:Biggest issue is still liability

[TheTurtlesMoves]

for me the biggest problem with self-driving cars is legal liability.

This is already covered. Brakes fail, tires blow out, mechanical failures happen. They kill people. Its been something that has happened plenty and gone trough the courts many times. Precedent has been set.

Re:Biggest issue is still liability

[bws111]

Seriously? First, your liability does not end where your coverage does. If you are under insured, it is you who us responsible. Second, the insurance company does not pay out of the goodness of their hearts, they pay because you pay them to. And if you have a claim, you will pay more.

The lower insurance rates with dash cams are more about fraud detection than dispute resolution.

Re:Biggest issue is still liability

[mjwx]

So, disregarding how the self-driving car decided who it is best to kill in any given situation

This old chestnut needs to die.

Rules for this already exist, it's just that human drivers dont follow them. An autonomous car will be programmed to take the course that causes the least damage and is the most legal. So they would choose a rear end crash over a right angle crash because a rear ender presents the lowest risk of casualties. If you think that veering out of your lane to avoid a rear end crash is a good idea, you completely fail at driving (first in failing to keep a safe distance, secondly in not having the time to check if it's safe to change lanes).

This is already sorted. They dont care that the person to the left is a meth addict and the person in front is a Nobel laureate, they'll take the path that presents the least risk. Beyond this, if the car is already in this situation the programming has failed.

If Google wants to sell autonomous cars, Google should be liable for anything the damned thing does.

So Colt and Armalite should be made responsible for mass shootings because they make guns?

I think firearm restrictions are sensible, but even I have to admit that the problem isn't the people who sell guns, it's the people who use them.

So once again, this is a legal problem that is already sorted. We already have extensive rules to determine who is at fault in an accident and they dont need to be changed for autonomous cars. When it comes down to it, if the driver cant be identified the onus comes back to the vehicles owner.

If you crash your car into a house and do a runner, you cant claim you weren't the driver because you weren't found at the scene. Unless you can nominate another driver who admits responsibility, it will be considered your fault (so be aware who you lend your car to). It will be the same with autonomous cars. For decades after their introduction you will still be required to pay attention to the roads, even though the car is choosing the actions you are still responsible for ensuring the car is doing its job properly and ready to take over if it fails. If it crashes whilst you were managing it, you will have failed your duty of care.

Re:Biggest issue is still liability

[mjwx]

Same thing that happens when a modern car with brake assist rear ends an old car with better brakes and traction.

If your car has shitty brakes you leave extra room. Good drivers realize that 'shitty brakes' is always relative.

If you're driving a crazy high performance car you moderate your brake use to avoid being rear ended.

This.

You also dont have to be driving a crazy high performance car to get good braking. Just get some performance pads, rotors, good tyres and maybe some braided brake lines and you can make a Toyota Corolla stop like a sports car. Your 0-100 time will still be crap but 100-0 will be amazing. You dont even need to fit six piston callipers.

You've got to understand your car. Sadly this is something most people never learn. They get in it every day but dont understand where the edge of the envelope is. When I get a new car, I take it to an empty car park and test it out. Most of this testing is about 1/2 an hour of practising parking to make sure I get it right, but I also test braking, turning and accelerating to get used to how the car behaves. I know it sound very yobbish, but it's actually rather sedate with very little tyre smoke. I normally only do one emergency brake test from about 40 KPH (~25 MPH).

Also the first thing I usually upgrade is the brakes. I've currently got Project Mu pads, DBA rotors and APP brake lines on my S15, I did these before getting a bigger turbo. I've thought about getting the 6 piston callipers off of a Skyline (they'll easily fit on my Silvia) but its not really worth it unless you're turning into a hardcore track car.

Re:Biggest issue is still liability

[Anonymous Coward]

Maybe that's why my parents never bought me a car or let me drive.

Re:Biggest issue is still liability

[sl149q]

And that is why we have insurance.

Why do you assume that insurance would not be available? And if insurance is available why is this an issue?

Sure if there are multiple vehicles and multiple insurance companies they might have a proxy battle to determine the answers to these questions.... but that is how issues like this have been sorted out for the last couple hundred years and will continue to be sorted out. Once you have precedents the insurance companies just adjust their rates accordingly.

Re:Biggest issue is still liability

[sl149q]

Exactly. Do you think you ARE NOT paying insurance when you are in the back of a taxi? Its just that the fare reflects the operating cost. And part of the operating cost is the insurance. And you really want the driver to HAVE insurance. So you pay the fare which pays the insurance.

How is that different from buying insurance to cover the self driving car you buy or lease or rent to get you from point a to point b. The insurance is there to make sure that any parties injured in a collision (including yourself) will be covered.

Re:Biggest issue is still liability

[Neil Boekend]

That computer would stop on the on ramp of SF bay area highways and refuse to move.

Why? Too steep to brake? Too many tailgaters?

Re:Biggest issue is still liability

[Anonymous Coward]

Why do you assume that insurance would not be available? And if insurance is available why is this an issue?

Sometimes when a car crash causes enough damage, we as the society are not content to get the payment from the insurance companies but we also send the responsible person to prison for it.

If the law says that I am the responsible person even though I don't have any real possibility of preventing the accident at all, I'm not going to get in the car. Even if the car manufacturer assures me that with their software that can't happen.

Re:Biggest issue is still liability

[Neil Boekend]

Dunno how it is in the US, but here in the Netherlands you are obligated to prevent parts from dropping off your car. Also, the quite thorough yearly tests check for such cases.
The first car would be liable.
In practice it is impossible to find liability in such a case. If I drive behind a car that looks like it's going to loose parts I'll keep an appropriate distance. It doesn't happen all that often.

Re:Biggest issue is still liability

[dave420]

The computer is in a very good position to know when those components develop faults, as it knows with far greater precision when it takes more effort/time to perform some function (turning, braking, accelerating, etc.). Monitoring the other components is also trivial, and a computer can do a much better job than a human.

Re:Biggest issue is still liability

[HornWumpus]

Too many tailgaters basically.

Re:Biggest issue is still liability

[HornWumpus]

Your performance is limited by the tire contact patch vs the weight of the car. (Don't even answer if all you've got is the first order friction approximation.)

Upgraded brakes without upgraded tires first is almost certainly a waste of money. Were the old brakes incapable of locking the tires?

Brake upgrades will improve your control and brake feel. But the tire is the limiting factor.

Re:Biggest issue is still liability

[Plumpaquatsch]

Welcome to a world with lawyers and liability laws.Someone is always to blame.

And, as I said, you can bet your ass Google et al are going to try to make sure it's you and not them.

So? Don't drive a Google car then.

Even if that means you'l be in at least 10 times as many accidents where you are at fault as a consequence.

Re:Biggest issue is still liability

[Plumpaquatsch]

Cars are very complex machines that can have loads of things go amiss with them without rendering them undrivable, and can have loads of other things go wrong while in operation. You're correct that it's safe to assume everything will be recorded, but I expect equipment failures will plague first-generation autonomous cars once they're old and the tolerances have loosened up. Steering, tires, brakes, suspension alignment, all things that will lead the computer astray as it's attempting to self-drive.

If the computer still works, it will realize that the parts don't operate at full capacity anymore, and will drive to account for that - including refusing to drive when it can't do it safely. Including when self-analysis shows itself isn't working anymore. Just like cars today will stop or only allow driving at low speed when they detect a problem.

Re:Biggest issue is still liability

[mjwx]

Your performance is limited by the tire contact patch vs the weight of the car.

And the contact patch between the brake pads and the brake rotors.

You may as well just say "I know nothing about cars" as that would have been a shorter explanation.

Brake pads and rotors of better design and materials are able to work better at both lower and higher temperatures. Yes, cold brakes do not work as well as warm brakes (and we'll ignore brake fade as you shouldn't be getting that on a street).

Beyond that, with modern ABS, your brakes should not be locking up your tyres with ease, leaving a lot of room for improvement. Seeing as friction is also the force acting between the brake pads and brake rotors you cant argue that improving tyres will improve braking performance whilst improving brakes will not.

This is no moral decision

[prefec2]

Humans are unable to make moral decisions in a few miliseconds. They would either freeze for a least one second and hit the next car or pedestrian depending on which comes first. If they have more time, they would try to avoid collision with the human and hit the car, because you cannot really see other people in there and you do not know how many persons are in there. Also people in the car are better protected. So the safest thing is hit the car. But beside that people know when approaching an truck trailer and they cannot stop, they should aim for the wheels and not the section in the middle. However, most people are unable to implement that so why should be cars be able to do these things?

Re:This is no moral decision

[Registered Coward v2]

Humans are unable to make moral decisions in a few miliseconds. They would either freeze for a least one second and hit the next car or pedestrian depending on which comes first. If they have more time, they would try to avoid collision with the human and hit the car, because you cannot really see other people in there and you do not know how many persons are in there. Also people in the car are better protected. So the safest thing is hit the car. But beside that people know when approaching an truck trailer and they cannot stop, they should aim for the wheels and not the section in the middle. However, most people are unable to implement that so why should be cars be able to do these things?

You have hit on one of the key reasons why trying to implement human reasoning in an emergency; especially since it's usually a subconscious reaction to avoid hitting the bigger, scarier thing. yo can train people to make calm decisions in an emergency situation but that takes a lot of simulator time and practice; something most drivers sorely lack before getting a license. If you wanted to follow the human reasoning it would simply be "CRAAAP.... AVOID HITTING THE BIG THING...DAMN... A PEDESTRIAN ... OH WELL IT ISN'T THE BIG THING...."

Re:This is no moral decision

[Livius]

I would go further and say that those who believe that ordinary people think rationally and/or ethically in the spit second of crisis in an imminent collision are possibly sociopaths and should not be allowed near critical software that will these kinds of decisions.

Don't like driverless cars?

[Anonymous Coward]

Put Andreas Lubitz behind the wheel.

Regulations?

[Coren22]

As opposed to all the laws and regulations making driverless cars difficult to test in the US? Google has to pay someone to sit in the front seat so they can take over from the computer (that can make better decisions faster than a human).

What regulations concern them so much (I didn't see any listed in the article) and how do they differ from the US regulations (like the US in some lawless state..)

Re:Regulations?

[prefec2]

The thing is quite simple. They have tested such cars in Germany. Therefore, it is possible to do so. They can also test their shiny new thing in German traffic, as long as a human back up is sitting in the car. And testing it in a German or other European inner city should be challenging enough for now. However, they want to show how innovative they are and not get surprised by the Japanese again, as it was with the hybrid. So actually this is mostly advertisement.

Love how they avoid the things humans CAN NOT DO

[gurps_npc]

The anti-driverless car always love to bring up the situation that they think the human handles well, but the computer does poorly - the given example of hitting a pedestrian vs a family in a car. This of course ignores the fact that 99% of the time humans have no idea if the car has a family in it, or a single neo-nazi.

But the self-driving cars ARE capable of hitting the breaks quicker and more reliably (avoiding skidding) than a normal human would

Think about it if it were the other way around - what if humans were crappy about deciding to hit the pedestrian but computers had incredibly slow reflexes and took ten times as long to decide to hit the break. Given that example we would laugh and say no way would we let anyone with slow reflexes drive a car.

But we already do that - we let human reflexes drive a car - (Even if they have had one drink 30 minutes ago, slowing them down). The question is not and never has been will computers be perfect drivers. Instead the question is will they do

• better than humans in most situations.

And that is something that we likely can do within the next couple of years, if we can't already do that.

So stop being obstructionists idiots bringing up the rare/never seen in the real world situations, and talk about what actually happens.

If all they want to do is test

[Anonymous Coward]

.. they can still put a human behind the wheel take actions in dangerous situations.
Or they could run their tests in the U.S.

Bullshit

[Skylinux]

For example when faced with the decision to crash into a pedestrian or another vehicle carrying a family, it would be a challenge for a self-driving car to follow the same moral reasoning a human would in the situation. 'Technologically we can do fully automated self-driving, but the ethical framework is missing,' said Volkswagen CEO Martin Winterkorn.

We learn how to drive in driving school and not how to crash. In a situation like the above. Most people will hit the next thing in front of them, regardless of what or who it is.
The faster the car goes the less likely anyone is to avoid an obstacle.

If anything. A machine could improve things by being able to react in time.

Re:Bullshit

[citizenr]

We learn how to drive in driving school and not how to crash. In a situation like the above. Most people will hit the next thing in front of them, regardless of what or who it is.

I remember stats showing drivers going into a tree at high speed are more likely to aim (unconsciously) for the passenger side - drivers have higher chance of survival than front passengers.

>> same moral reasoning a human would in the situation

This is some hilarious shit right there. There are no higher mental functions involved in a crash, its all on instinct, pre-learned behaviour and reflex.

Quite a high standard compared to us humans who wo

[Anonymous Coward]

Seems someone thinks we are rational actors while swerving to avoid death . Maybe best is to apply brakes at point of least collision while coordinating with other nearby cars and call it a day ?

The Germans just need to do development in the US

[deck]

While they are trying to change the laws back home they might as well do their development and testing in the United States. We currently have fewer restrictions here.

I agree with gstoddart about autonomous cars being able to be 100% hands off by the user at all times for normal driving regimes. If the companies that make them do it right then they should not be afraid of being 100% responsible when the vehicle is in autonomous mode. Some computer modeling of autonomous vehicles has shown a major drop in the number of accidents; hopefully this can be realized in real life. However, owners will have to have insurance to cover the vehicle as property and when they are operating it.

Re:Show me 1 independently publicly tested autonom

[afidel]

ahem

Re:Love how they avoid the things humans CAN NOT D

[friedmud]

I somewhat agree... but the problem is not "is a driverless car better than a human" it's: who do we sue when something goes wrong.

In the proposed hypothetical, whoever gets hit is going to be suing someone. Who do they sue? The owner of the car (even if they weren't in the car the time?) the "passenger" or the company that makes the vehicle.

I tend to think that it will be the owner - and the owner will need to have insurance to cover whatever the autonomous car could do. There is no way a company like Google is going to put out a product like this where it's liable.

As for "rare/never seen in the real world situations" you have to remember that with a high enough saturation of these on the roads... those "rare" situations will happen all day long, every day (they already do with regular cars). The statistics makes this quite tricky...

Note that I am NOT an obstructionist and I am all for autonomous vehicles... but the legal aspects of self-driving cars is something that needs to get solved soon...

But

[invictusvoyd]

'Technologically we can do fully automated self-driving, but the ethical framework is missing

Ethically we can allow fully automated self-driving, but the Technological framework is kinda missing

your car may kill you to limit casualties

[Anonymous Coward]

Should car hit mom crossing a street with twins in the stroller or three retired older people on the side walk?
Should it hit Einstein or a group of dozen people?
By the way, one of the options will be your life as a driver of a new smart car. Your car may calculate that your life is worth less than others.

So how do you decide? least amount of casualties? life expectancy of victims? number or surviving relatives? annual salary/wealth? importance to society?

Without lawyers defining ramifications of what ethical car should do we will end up with a company releasing a code liable for deaths of people.

Morality Framework UNNEEDED

[DumbSwede]

Why this obsession with moral reasoning on the part of the car? If using self-driving cars are in 10x fewer accidents than human driven cars, why the requirement to act morally in the few accidents they do have. And it isn’t as if the morality is completely missing, it is implicit in not trying to to hit objects, be they human or otherwise. Sure try to detect which are objects are human and avoid them at great cost, but deciding which human to hit in highly unlikely situations seems unneeded and perhaps even unethical in a fashion. As it is now, who gets hit in these unlikely scenarios is random, akin to an Act of God. Once you start programming in morality you’re open to criticism on why you chose the priorities you did. Selfishly I would have my car hit the pedestrian instead of another car, if the latter were more likely to kill me. No need to ascertain the number of occupants in the other car. Instinctively this is what we humans do already -- try not to hit anything, but save ourselves as a first priority. In my few new misses (near hits) I’ve had, I never find myself counting the number of occupants in the other car as I make my driving decisions.

Re:Morality Framework UNNEEDED

[trekdvorak]

Might be a generalization. Given the option, some drivers may steer themselves towards danger to avoid endangering others on the road. I for one though would not want this decision to be handed to a computer as I believe it is mine to make. This transcends morality and begins to touch on a human rights issue. IMO

Re:Morality Framework UNNEEDED

[drinkypoo]

Instinctively this is what we humans do already -- try not to hit anything, but save ourselves as a first priority. In my few new misses (near hits) Iâ(TM)ve had, I never find myself counting the number of occupants in the other car as I make my driving decisions.

It's a horribly stupid breakpoint in any case. The truth is that if you swerve to avoid a school bus and run over an old person on the sidewalk, they're just going to do you for running over the old person because you weren't supposed to be on the sidewalk. Meanwhile, you weren't supposed to be outdriving your vision or your brakes, so you wouldn't have had to dodge the school bus anyway. The self-driving car won't outdrive its vision or brakes, so it won't have to swerve. If someone jumps in front of it, it's going to decelerate slightly right before it smacks them into next week, and the onboard video will prove that the car could not have stopped so the pedestrian is at fault.

Re:Morality Framework UNNEEDED

[DumbSwede]

Ahhh, but you are looking at the one situation in isolation. The moral thing to do is everyone hand over the driving to the machines as that will save the greatest number of lives in the long run. By being unwilling to hand the decision to a machine you are choosing to kill a greater number of humans in practice on average – just so you can exercise the moral decision in some outlier. If self-driving cars were only as good as, or even possibly just a little better than us at driving, I might side with you, but likely they will be orders of magnitude better.

BTW I meant “former” not “latter” in my first post.

Re:Morality Framework UNNEEDED

[sl149q]

While you might like to make the decision in most cases there simply isn't time in the fraction of a second you have available during a crash.

The issue about moral decisions with self driving cars arises because for computers, first they will have far more situational awareness. They will have been monitoring and possibly making worst case projections showing the possibility of an accident for a long time (for computers seconds is a very long time...)

So even once the probability of collision reach 100% there may be decisions that can be made based on previously collected information and previously projected possibilities. And there may be lots of time to influence the outcome with differential braking, steering etc.

So AT THAT POINT there needs to be rules that govern how to make those decisions. Humans cannot because they are not fast enough in most cases. Computers can and must make decisions. Those are predicated on algorithms written by programmers. So there needs to be a basis for making them.

Re:Morality Framework UNNEEDED

[lucien86]

Assuming the machine is based on a Strong AI type system. -
A Strong AI is a dynamic real time type system, and has a lot of constraints that limit what it can do.. The underlying speed of the hardware might be very fast but its very hard to really make use of most of it. The machine uses a predictive algorithm that is normally looking over many seconds forwards and backwards in time. If a Strong AI machine gets into a disaster type situation there's probably not really a lot more it can do than a person because the probability cone is broken. Its main goals are likely to be simply bringing the vehicle to a stop and avoiding obstacles. Any priority choosing of where to go would intrinsically choose the machine and its passengers first, because the machine actually 'works' by continually optimising its own survival probability cone. If the car protects itself it intrinsically protects its passengers as well..

At first glance Weak AI is a totally different story. Here the algorithms are probably mostly fixed and pre-programmed so there is more scope for prioritising who to protect first in an accident.. The other side of this though is that the speed of a disaster situation may simply overwhelm the machines visual model leaving it effectively blind in the time it has. Again all the machine can really do in that position is try to stop as quickly as possible and maybe avoid obstacles.
A big difference is that weak AI machines may be restricted to fairly slow speeds whereas the Strong AI variant can probably drive safely at higher speeds than most humans.. (50 mph verses 200 mph) Strong AI based designs are probably still 15 to 20 years in the future..

Re:Love how they avoid the things humans CAN NOT D

[prefec2]

This is totally true. No human ever was able to make a moral determination and act on it in an accident situation. Beside the fact. The pedestrian and the family in the car is quite simple. The family is protected by their car the pedestrian is not. It would be a different thing between one or a group of pedestrians. Most humans would freeze and hit who ever comes first. End of story. The car could try to reduce victims.

Re:Love how they avoid the things humans CAN NOT D

[HornWumpus]

As soon as any autonomous car advocates start talking about 'what actually happens' the conversation can start in earnest.

But for now, all we have is Google's marketing BS and some DARPA challenges that paint a much less rosy picture.

Um, this is an easy decision for software to make

[Anonymous Coward]

Obviously, when it come to hitting a pedestrian or another car, regardless of how many passengers the car has: hit the car. At least the passengers in the car have a chance to survive, but if a car hits a person the chance for survival is much less. Especially, if that car is a large vehicle, or even a semi-truck. Either way the car passengers might survive, but a pedestrian is going to be more injured or dead. Dead for sure if he/she is hit by a semi-truck.

Re:Show me 1 independently publicly tested autonom

[Anonymous Coward]

Oh, sorry, I forgot to add "successful". These are very interesting research challenges, but nowt here about succeeding on an actual real-world road.

Already solved.

[redwraith94]

Self preservation should always triumph. Why? That is what people do anyway.

Re:Already solved.

[gnasher719]

Self preservation should always triumph. Why? That is what people do anyway.

That is obviously wrong. Minimising total damage should always triumph. If you intentionally drive into a pedestrian, you should and will go to jail.

Re:Already solved.

[Anonymous Coward]

I'd rather go to jail than a grave.

Pedantic insight

[Anonymous Coward]

Its not who the injured party will sue, its who will their insurer sue for them.

But you actually point out the biggest impediment to self-driving vehicless, the strawman of who will get sued in the event of accident. The software maker/car maker of course should be sued if they are at fault. If there is no fault with the software, it acted as it should have but the situation was the problem in the first place then someone else should be sued. In the example above the pedestrian IS the problem. This situation only comes about due to jaywalking in dangerous situations i.e. person steps out from between 2 vehicles that obscured them completely from the nearer driver from a distance. The pedestrian who created the dangerous situation in the first place should be sued, but good luck finding them to sue. That's the situation right now with manual driving already, the real at fault party is rarely around to get sued already (either form leaving the scene they caused or, you know, being dead). But we don't force cities to enclose all street sides to prevent jaywalkers by force based on the fact that the city might get sued if they don't!

To go with self-driving cars we need limits on their use. For example they could be made to only "take over" once the vehicle is on streets with greater than 25mph speed limit in place (this is a standard for my state's neighborhoods, your state/country may have different limits). Elevated or buried pedestrian crossings so they have paths to travel that do not cross traffic at all. Proper bike lanes with programming made to defer to them. and on and on.

The problem is not self-driving vehicless, the problem is everything else that MUST be upgraded to keep up with them and make the roads safer for everyone even without SDV's on the road.

Why American companies have it easier

[SuperKendall]

The summary doesn't really explain why that dilemma is harder for German companies to solve than American companies.

For Americans, the answer is: always hit the pedestrian(s). What the hell was anyone doing outside of a car?

Re:Show me 1 independently publicly tested autonom

[hatemonger]

If your requirement for caring about emerging technology is that it has already achieved commercial success, you're going to have a very short list.

Re:Love how they avoid the things humans CAN NOT D

[Qzukk]

As soon as any autonomous car advocates start talking about 'what actually happens'

Why yes! Just the other day a baby stroller magically appeared 2 feet in front of me while I was doing 90mph on the local autobahn, forcing me to make a snap decision between creamed baby or ramming the President's car which was carrying a gaggle of pregnant neurosurgeons to a peace conference that just happened to be in the other 5 lanes of the freeway and the shoulders and the sidewalks and the ditches, all at the same time (except for the lane the baby stroller was in, of course).

    Fortunately being the superior human being that I am, I was not constrained by mere binary thinking and through a little ingenuity and physics, managed to lean my manly, manly physique across the vehicle just in time to ski past the baby and land safely on the other side. Not only that, but as it gracefully settled back onto four wheels, the breeze from the air displaced by my HMMWV (only the real ones are worthy of my leisure time) blew out a forest fire that was threatening to ignite a fireworks factory right next to a nuclear waste storage shed.

Let's see one of those auto-no-muss cars do that!

an EULA will not stand up in a criminal court

[Anonymous Coward]

an EULA will not stand up in a criminal court and I don't think that a EULA will stop a 3rd party victim from suing as well. Much less one they can't even give a ok to.

ethics

[Tom]

For example when faced with the decision to crash into a pedestrian or another vehicle carrying a family, it would be a challenge for a self-driving car to follow the same moral reasoning a human would in the situation

Or maybe it would follow better moral reasoning. Ours is not perfect, it's just whatever evolution came up with that gave us the best species survival rates. That doesn't mean it's really the most ethical solution.
For example, in a post-feminist society, let's assume for arguments sake that gender discrimination has been overcome, wouldn't we also do away with "women and children first" - which is a suitable survival approach in a species fighting for survival in the african prairie, but hardly for the dominant species that already is overpopulated.

Re:ethics

[Anonymous Coward]

In theory it should be children last, because the adults have had years of valuable experience, you can always make another kid and train them to the same level in a much shorter period of time.

Re:ethics

[Tom]

Correct. Rational evaluation will lead you to this result.

We're not rational, because our genes (who drive our evolution and thus our minds) don't give a fuck if we survive or not, only if they survive. To them, your kid is more valuable then you are.

Sit back and watch life live for you..

[Anonymous Coward]

Life is dangerous, deal with it and move on. It could be argued that removing the freedom to drive off of roads or to a destination of your own choosing is worse than the safety gained.

Re:Love how they avoid the things humans CAN NOT D

[HornWumpus]

Do you have a point? Besides the one on your head?

Easy Out

[Anonymous Coward]

Let the NSA have a real-time feed from the cameras on every self-driving car and the laws will be changed in a year to fill the streets with self driving cars. Morality will have been removed from the equation.

Re:Easy Out

[trekdvorak]

Although slightly over the top, you do bring up a good point about privacy. This would be the most ubiquitous form of surveillance...ever. Look around you, do you see a car? It sees you too.

Re:Show me 1 independently publicly tested autonom

[cayenne8]

For example when faced with the decision to crash into a pedestrian or another vehicle carrying a family,

That's an easy decision......smush the pedestrian, that will cause less damage to me and my car.

Driver vs. Robot Ethics

[jacksdl]

"For example when faced with the decision to crash into a pedestrian or another vehicle carrying a family."

Ethical human traveling 60kph when pedestrian and family in car simultaneously appear as obstacles 30 meters ahead..."OMG, OMG, I'm gonna hit one or the other! The car is better protected, but there are more people in the car. What did Spock say about the needs of the one? OMG " BANG!

Robot driver in same situation... "Obstacles detected, apply maximum ABS braking". Car stops 8 meters short of collision.

For every "ethical" dilemma a human decides correctly, I would guess there are 1000 fatalities caused by humans just being lousy at driving.

Oh, and what is the right ethical decision to make when deciding whether to hit a single pedestrian or a family in a car?

Re:Driver vs. Robot Ethics

[Anonymous Coward]

Oh, and what is the right ethical decision to make when deciding whether to hit a single pedestrian or a family in a car?

Pedestrian on a sidewalk, or on the road? Well, you can assume any vehicle on the road has consented to being in a dangerous situation. To go off road to hit a pedestrian seems like a poor choice. Just like pushing a fat person down a hill to save a train full of people, when the train full of people all consented to be at risk while the fat person did not.

Manufactured controversy

[flopsquad]

1) Cars are not technologically at a point where they have omnipresent awareness of the constituents of every vehicle around them and the locations of every pedestrian (add in crowded street-facing cafés, structural members for buildings, and everything else you could possibly think of). Neither, for that matter, are people.

2) The most brilliant philosophers still disagree over the ethics of choosing who dies when someone's gotta go. See also the Trolley Problem, most other ethical dilemmas, and generally the eternal struggle between various consequentialist and deontological systems of ethics.

3) This precise scenario is highly contrived and seems (1st approximation) to be vanishingly rare.

Given the above, maybe the question shouldn't be if a robot can make a perfect (or any) ethical decision. Maybe for now it should just be if the robot can do better than a human at not killing anyone at all in these sorts of situations. Maybe "I did my best to protect my owner from death and just happen to average out safer for everyone" will have to be ok for now.

Manufactured straw computer controversy

[burtosis]

The whole ethics framework debate is a straw man (computer) argument. It's patently obvious people don't make those split second judgement calls. The real reason the Germans are on sound moral grounds is autonomous cars are nowhere near commercial prime time on sunny day clear traffic straight highways. Dirty sensors, unpolished code with bugs, proper reliable extraction of features, sensor failures, intelligent prediction of object locations, prediction and proper avoiding of road hazards, and many more things still need to be ironed out. Problems creep up where a driver may have only a second to take over after the AI bails - complete bullshit that a human can hyper concentrate over the controls and take over in one second properly three hours and twenty six minutes in.

How the fuck do you justify putting people's lives at risk with your crap box wanna be AI? I want self driving cars as much as anyone not employed in the commercial driving industry. But not when any jackass car with half capable systems risks everyone's life. Honestly it needs to be banned in the United States on public roads. Wanna play? Pony up for a closed course before risking people's lives. If you think moral decisions are the barrier to AI cars you haven't the foggiest idea what the actual challenges are.

Re:Manufactured straw computer controversy

[flopsquad]

Not sure if the "you" in your post was me or the Googles of the world making self-driving cars. If it's me, I'll just point out that I never proposed that handwringing over decisional ethics was the one thing holding SDCs back.

My point was that questions like the one in TFS are matsurbation. The question ought to be, are we at a point where they're safer (aggregate) than humans, driving in real world conditions? You and I both agree that currently the answer is no. For optics and liability reasons, they'll probably have to be an order of magnitude safer than the average human driver before they gain wide traction. I think that day is closer than you seem to, but that's just fortunetelling.

I disagree about rushing to a blanket ban, and I don't grok your main complaint about jackasses with half capable systems. Is there a big mod/DIY community out there outfitting their Suburbans with hand-rolled CarLinux or something?

FWIW I agree the whole "don't worry the driver's gonna be right there to take over at a moments notice" line is absurd. I buy it for these early test runs where the drivers are paid to make sure no one gets flattened by a Google logo, but a real self-driving future actively discourages engagement and driving skill.

Re:Manufactured straw computer controversy

[burtosis]

I didn't mean to imply you but in general.
I pretty strongly object to testing in real life situations when the populace has an expectation of safety. The google car is perhaps the most advanced in the world yet is not able to function safely in city driving. Google themselves admit it's not ready or it would be rolled out as a product. It's a far cry from teams of engineers, programmers, and scientists fussing over every last detail, planning routes where only expected problems (if any) are in ideal situations and real life situations where these vehicles will be dirty, neglected and abused.

Personally i dont feel that its reasonable to allow any self driving cars on any public roads until they have rigorous safety testing completed on closed courses first. I'm less worried about google than some of these fly by night companies that really are close to a mod/DIY community level of technical ability. The latest test by Delphi comes close to that.

Ethics are an interesting dilemma

[Golden_Rider]

Let's say they manage to program the car so that it can calculate which course of action will cause the least injuries/fatalities. Now you get into a situation where the only two options available are a.) evade some obstacle on the road, but thereby hit a group of five pedestrians, quite possibly severely injuring or killing them or b.) hit the obstacle, quite possibly killing the driver (you). You are alone in your car.

Now, would you drive such a car which sometimes can decide, with cold, pure Vulcan logic, to kill you?

Re:Love how they avoid the things humans CAN NOT D

[Anonymous Coward]

I somewhat agree... but the problem is not "is a driverless car better than a human" it's: who do we sue when something goes wrong.

In the proposed hypothetical, whoever gets hit is going to be suing someone. Who do they sue? The owner of the car (even if they weren't in the car the time?) the "passenger" or the company that makes the vehicle.

Yes. When you file your lawsuit you name all the possible entities... then you let the court decide how to apportion the blame.

That ethical challenge is nil

[aepervius]

Those made up example with the family van and the walker are laughable. Firstly out of my experience you do not have time while being in an accident to think that far as to check & see the family van is full of children. You steer , brake, and do as much as you can to avoid the walker. And the van. Secondly you really think there is even a concurrence ? The walker will get the full hit of the kinetic energy in his body. The van will absorb part of it in its structure. What sort of SICK FUCK would hit the walker rather than the van because there are children in the van ? There is no photo finish : You steer to avoid all, and if you cannot, you try to hit the target which will get the less physical damage from your damn car *if you can think that fast*. Does not matter if the walker is an old 99 guys and the van full of the brim of puppies and babies. I expect any automated car to make the same damn calculation : between hitting a car and a walker, go for the car. Hit the one with the most protection.

Re:That ethical challenge is nil

[Anonymous Coward]

A car will never make those decisions until we have strong AI. The car doesn't know if a family is in the van. The car doesn't know if the van has crumple spots or not. The car knows there's a thing over there with a specific movement vector and another smaller thing with a different vector. One is behaving like a walker and the other like a vehicle, but can't confirm or deny if those objects are actually those things.

When the car believes it's going to hit something it'll brake hard. It's not going to go driving all over the street and sidewalks at 30 like in the movies. It'll stop in it's lane or it'll hit. It'll probably stop and miss as it was following the van at a safe stopping distance, something few humans do.

Or are all the auto manufactures and large tech companies sitting on massive sensor and AI advancements? I think they'd make more profit selling those things instead of trying to be the first with an automated car.

Re:That ethical challenge is nil

[Delgul]

Hmm... Call me a sick fuck then because rather than killing myself, my wife and my children driving into that upcoming 2 ton truck, I would choose to hit the pedestrian every day of the week! And to add a bit of fuel to the discussion, I (and I suspect there are many that agree with me) would only ever buy a car that makes decisions that takes ME into account first and THEN starts looking at minimizing collateral damage. But, like you said, I'm a sick fuck by thinking this apparently...

Re:That ethical challenge is nil

[dave420]

I'd rather call you a sick fuck for driving beyond the range of your vision and braking, as that's endangering your family every single time you drive them. A safe driver would never voluntarily get in a position where they have to make such decisions - they'd give themselves enough distance to brake for anything that might happen. Where did you learn to drive? They didn't teach you too well.

Re:Love how they avoid the things humans CAN NOT D

[mjwx]

The anti-driverless car always love to bring up the situation

The pro-driverless car crowd always love to ignore the fact that the autonomous car wont be driverless for decades. A human will still be required to oversee and in case of a failure, take control of the vehicle.

The big problem with this is that people will be taking manual control because the autonomous car will abide by the rules that human drivers like to ignore like keeping a safe distance, not driving in the passing lane, keeping to the speed limit and slowing down in potentially hazardous areas (I.E. roads frequently entered by pedestrians).

Human nature wont change overnight because a pro-tech crowd wills it.

How about

[mrprogrammerman]

not hitting anything. Unless you're going really fast, I would expect the longest part in stopping a car from hitting something is the human reaction time. Take that out of the equation and driver less cars shouldn't be hitting anything.

Wrong logic

[snowtigger]

"For example when faced with the decision to crash into a pedestrian or another vehicle carrying a family, it would be a challenge for a self-driving car to follow the same moral reasoning a human would in the situation."

No, a self driving car shouldn't get into that situation in the first place. The right thing to do here is to anticipate events and slow down. Self driving cars have a huge advantage here, in that they don't get tired or lose attention over time.

Wait what question?

[DarkOx]

For example when faced with the decision to crash into a pedestrian or another vehicle carrying a family

Um there is no question for several reasons.

First if the situation is so immediate your only two options are hit a vehicle or hit a person its highly unlikely you have time to peer into the other vehicle and count its occupants.

Second most vehicles on the road today have lots of safety features; if they are being used, seat belts fastened airbags not disabled etc, most crashes are highly survivable; most pedestrian vehicle crashes far far less so for the pedestrian (excepting very low speed nudged someone in a parking lot cases).

Finally while your liability insurance should most likely be on the hook in any situation I can image you finding yourself faced with such a choice another driver is more likely to have supplemental coverage that will ensure they are taken if your insurer dicks out and tries to screw them.

I really can't image a situation, most things being held equal, ( I know you could contrive a situation where you will be nearly at a complete stop before you hit the pedestrian vs hitting the other vehicle at high speed ) where it would ever be appropriate to choose to hit the pedestrian.

Re:Love how they avoid the things humans CAN NOT D

[dave420]

But it can be legislated away if driverless cars are safer than their meat-driven alternatives.

Re:Not concerned / The real issue is security

[lucien86]

You are absolutely right about security. People just don't seem to have done the basic marketing on autonomous vehicles. - One of the biggest dangers for all autonomous vehicles - or autonomous machines in general is theft.
The only solutions are that machines are kept under constant human guard - and or - are set up to defend themselves against theft. One of the biggest problems is emergency stop buttons, or 'Steal Me' buttons. - The solution for Strong AI machines is that there is no absolute stop button but there is a priority override button. The machine will only stop if it recognizes an authorized user or owner.
Even there the maze gets complicated because the machine must also obey people like police or officials - but it must also recognise the difference between real and fake police. Maybe a solution is for people like police to carry special encoded radio badges, or to have the machines carry an on-board database of all authorised police & officials.

Re:Not concerned / The real issue is security

[TWX]

Around here most emergency responder vehicles have transmitters that can tell the traffic lights to change as they approach. It would not be hard to design such a system for short-distance transmission to alert autonomous vehicles to make way.

Re:Show me 1 independently publicly tested autonom

[Plumpaquatsch]

Oh, sorry, I forgot to add "successful". These are very interesting research challenges, but nowt here about succeeding on an actual real-world road.

https://en.wikipedia.org/wiki/Eureka_Prometheus_Project

The next culmination point was achieved in 1995, when Dickmanns re-engineered autonomous S-Class Mercedes-Benz took a 1000 mile trip from Munich in Bavaria to Copenhagen in Denmark and back, using saccadic computer vision and transputers to react in real time. The robot achieved speeds exceeding 175 km/h on the German Autobahn, with a mean time between human interventions of 9 km. In traffic it executed manoeuvres to pass other cars. Despite being a research system without emphasis on long distance reliability, it drove up to 158 km without any human intervention.