Volvo Will Accept Liability For Self-Driving Car Crashes (bbc.com)

An anonymous reader writes: Volvo has announced it will accept "full liability" for accidents when one of its cars is driving autonomously. It joins Mercedes and Google in this claim, hoping to convince regulators that it's worthwhile to allow testing of such vehicles on public roads. Volvo's CTO said, "Everybody is aware of the fact that driverless technology will never be perfect — one day there will be an accident. So the question becomes who is responsible and we think it's unrealistic to put that responsibility on our customers." Of course, this is limited to flaws in the self-driving system. If the driver does something inappropriate, or if another vehicle causes the accident, then they're still liable. It's also questionable how the courts would treat a promise for liability, but presumably this can be cleared up with agreements when customers start actually using the technology.

Comments

Easy to do when backed by the PRC

[sethstorm]

Given that Volvo is now a PRC-backed concern under Geely, it's easy for them to just throw out money.

Re:Easy to do when backed by the PRC

[Virtucon]

Don't be too sure of that, China has been burning through it's currency holdings lately to prop up the economy.

Re:Easy to do when backed by the PRC

[HornWumpus]

Shanghai stock exchange has an average PE ratio of 70. After 'the correction', It was 100.

Posting this information from China will get you arrested for talking down the market.

That's what happens when you have a real society where the 1% can have the government prop up their bad investments.

China has for decades been managed for a single metric...100% industrial utilization. It might turn out that wasn't such a good idea.

We can only hope the Chinese finally remove the peg from the Yuan. So far we only see it move rapidly when it helps them.

The Shanghai stock market is still at least 4x overvalued and it is not being allowed to move. Keeping that up will require infinite money. Something will give.

Re:Easy to do when backed by the PRC

[HornWumpus]

In about 40 months China will have spent it's entire net currency reserve from the last 30 years. I don't expect it to continue that long, but even China can't bleed $100Billion/month.

Re:Easy to do when backed by the PRC

[kheldan]

In about 40 months China will have spent it's entire net currency reserve from the last 30 years. I don't expect it to continue that long, but even China can't bleed $100Billion/month.

Oh, don't worry for China. When they're out of money an in danger of their economy collapsing, they'll probably just let North Korea off it's leash and let them start an armed conflict with someone, probably South Korea, to draw the U.S. into a war.. which then China, naturally, will have to jump into to defend it's friend North Korea, and WW3 will begin. Everyone knows that war is great for an economy, right? That'll solve all their problems.

Re:Easy to do when backed by the PRC

[lgw]

This is why my only China investments are tech stocks. China's internally-focused heavy industry is mostly a sham, and the huge manufacturing base outsourced there is being replaced by in-country automation, but the tech stocks there are real companies, with P/Es that are pretty normal for tech stocks. It's the one place where a 70P/E looks good.

Re:Easy to do when backed by the PRC

[truck_soccer]

That shouldn't be too hard considering that money isn't a real thing.

Re:Easy to do when backed by the PRC

[Aighearach]

Well, it was easy for Google to do it first, so your comment has no insight value.

As to the summary, no they won't just make a general "promise" for the courts to work out, they'll simply issue insurance and the courts won't see any difference except which party is providing the insurance. It will still work the same way; you sue the owner of the vehicle, and the insurance company provides their lawyer and pays any judgement or settlement.

If it is a dual-mode drivable car, then it will be up the two insurance companies to look at the vehicle "black box" and determine if the driver or the computer was in control, and that will determine which insurance company handles the claim.

There are lots of new things here, but how the court will handle liability will not require any new parts. It will just be normal auto and insurance cases.

Re:Easy to do when backed by the PRC

[Altrag]

Things may change in 20, 30 or 50 years but at the moment, China would have to have their back right up against a wall before starting a war with the US. They might have superior manpower but that only helps if they can get their guys armed and on the ground, and much of modern warfare is fought in the air these days where the US still reigns supreme for the moment.

And of course, cutting ties with your biggest trading partner is never good for the economy.

I could see China letting NK off the hook in order to destabilize the region and see what they can claim of the fallout.. but it would likely be a bit of a last-resort since the US would back whoever NK attacks (most likely SK but with a strong possibility of Japan) and China would be hard-pressed to take that fight.. so releasing their hold on NK would effectively be equivalent to just letting it go and hoping the pieces fall in their favor after everything is said and done (which they might.. I mean if NK doesn't get annexed by somebody after such a war, China's still their best bet for a friend, even if the US forces in an ostensibly "democratic" government.)

Re:Easy to do when backed by the PRC

[HornWumpus]

Do you play civilization?

Do you realize just how big a size multiple S Korea's economy is vs the north's.

America has been in the DMZ to keep the south from going north for 20 years or more. It wouldn't be much of a fight. Only the nukes and China complicate things.

Sure N Korea has lots of old artillery. S Korea has the gun positions zeroed.

Re:Easy to do when backed by the PRC

[HornWumpus]

If their regular stocks average 70, what do their tech stocks average?

Also how can you trust their books? The only way to make them less trustworthy would be to have Arthur Anderson do the audit.

Re:Easy to do when backed by the PRC

[lgw]

I don't trust their anything,but that's no different from US tech stocks. Speculative investments have a small place in my portfolio, and oddly the Chinese tech stocks have fared better than my biotech stuff in this downturn (though the latter had been up more).

Re:Easy to do when backed by the PRC

[kheldan]

No I don't play games or really care about the facts, I just felt like making sarcastic remarks about China and their BFF Best Korea. China bullshit puts me in that sort of mood. Their economy will tank, and they'll blame the West for it somehow. Meanwhile Best Korea keeps yapping and baring it's teeth like a chihuahua with rabies. Elsewhere in the world there are serious assholes running around blowing up archeological sites and cutting off people's heads just for the hell of it, and Russia is blowing up everyone but who they should be blowing up. All of it makes me a bit irate some days, you know what I mean?

Re:Easy to do when backed by the PRC

[HornWumpus]

I do have the 'bad feeling' about the outcome. Then again, foreign markets will buffer their crash.

I'd be even more worried if I believed the average Chinese citizen had much money in their markets or banks. I know enough Chinese people to bet they've got as much of their net as hidden as possible, in tangible assets, preferably farmland. Maybe it's a stereotype...They bought up all the rental houses in Sacto 2-3 years ago. (Farming fat kids?)

I don't think it will mean much

[Virtucon]

Even though they'll take responsibility, in every state in the US you must still have liability coverage. If these companies are to be their own underwriters so to speak then they'd have to jump through hurdles to be approved to operate as an insurance company as well. They could obviously partner with insurance companies as well.

Re:I don't think it will mean much

[CaptainLard]

If insurers only had to deal with 4-5 automakers vs 50 million drivers it would certainly reduce their paperwork and the hassle that goes with it. My friend is a claims adjuster and has a bunch of stories along the lines of: "Someone stole the oil out of my engine which is why my car broke down so I need you to pay me for it".

Anyway I'll file this under common sense. If Volvo software is driving your car and causes an accident, Volvo (the driver) is liable. If something else causes an accident involving your car that something else is liable. Just like we have now!

Re:I don't think it will mean much

[gstoddart]

But it will have to be made to mean something.

I've been saying for quite a while that self-driving cars can't just go into a failure mode which says "OK, meat sock, you do it I'm confused" and expect humans to be able to respond or take liability.

It's completely unrealistic to expect humans to transition from not actively driving to being required to take over in the event of an emergency.

Why would I pay insurance on a self-driving car? That would be idiotic, and basically means everyone else is footing the bill for the adoption of unfinished technology.

If the passengers aren't the source of the risk, they sure as hell shouldn't be the ones pay for the insurance.

Re:I don't think it will mean much

[tripleevenfall]

Volvo is offering to indemnifying individual owners against flaws in the self-driving system. Of course, you'd have to prove somehow that the self-driving system was responsible, and do it by going up against a massive corporation's legal department.

Re:I don't think it will mean much

[mh1997]

Even though they'll take responsibility, in every state in the US you must still have liability coverage. If these companies are to be their own underwriters so to speak then they'd have to jump through hurdles to be approved to operate as an insurance company as well. They could obviously partner with insurance companies as well.

In many states, you do not need insurance, but proof of financial security which can be a surety bond with the DMV. In my state it's only $50,000. The company I work for does this and we are in no way associated with or approved as an Insurance Company.

Re:I don't think it will mean much

[Nidi62]

Even though they'll take responsibility, in every state in the US you must still have liability coverage. If these companies are to be their own underwriters so to speak then they'd have to jump through hurdles to be approved to operate as an insurance company as well. They could obviously partner with insurance companies as well.

Or they just work with their current insurance providers. Sure they would have to renegotiate rates, but since major companies (especially manufacturers of finished goods) already have insurance for all types of liabilities both during and after production it should be relatively easy to extend said liability to driverless cars. No need to act as your own insurer.

Re:I don't think it will mean much

[gstoddart]

Sit in the back seat. Or don't buy one.

Honestly, until they get the issues of liability sorted out, the self driving car is a complete non-starter .. precisely because of crap like this.

Re:I don't think it will mean much

[tripleevenfall]

Volvo *seems* more trustworthy than most, and perhaps it would be your insurance company paying your claim, and then deciding to fight it out with Volvo themselves. However, what if they deny your claim? Are you going to be able to fight them? Will anyone?

Really, this gesture doesn't mean much. There will have to be a new structure devised for determining liability around self-driving cars. New laws, new types of insurance, etc. (I'm guessing it isn't going to be cheap to insure one of these)

None of that exists yet, but what Volvo is saying isn't much different from any other manufacturer's warranty, is it? They accept responsibility for accidents caused by design flaws and things that are "their fault" today in the form of recalls and whatnot in regular old meat-driven cars.

Re:I don't think it will mean much

[DigiShaman]

A vehicle won't be 100% autonomous, and least not for the foreseeable future. Meaning, there's still a steering wheel and peddles for human interaction. So until those are removed, expect to still pay insurance for the vehicle. Even still, you have other natural disasters that can total a vehicle while parked, hail, floords, landslides, theft...etc.

Most like these newer vehicles that employ autonomous driving will allow for auto insurance at a reduced rate for the owner. In fact, in the insurance business there's all sorts of factors that go into your rate such as the type of car, age of driver, location, and safety features based on make/model.

Re:I don't think it will mean much

[TechyImmigrant]

Sit in the back seat. Or don't buy one.

Honestly, until they get the issues of liability sorted out, the self driving car is a complete non-starter .. precisely because of crap like this.

This is for governments to make laws just as they have for vaccines. It's the same deal. Over a population it will save many lives. In some specific instances it make kill someone. The government gets to make laws that protect the vendors from the specific cases so the general case can be realized,

Re:I don't think it will mean much

[Jason Levine]

My guess is that self-driving cars will, in the case of accidents, have a "black box" that will be able to tell investigators just what was going on with the car including whether self-driving mode was engaged or not. So if the accident investigators determine that your car was at fault, but your car was in self-driving mode at the time, you'd be off the hook for liability.

Re:I don't think it will mean much

[eth1]

Volvo is offering to indemnifying individual owners against flaws in the self-driving system. Of course, you'd have to prove somehow that the self-driving system was responsible, and do it by going up against a massive corporation's legal department.

On the other hand, if your self-driving car has a crash, who do you think people are more likely to sue? You, or the corp with deep pockets? Lawyers will probably be lining up for contingency fees to go after the corp.

Re:I don't think it will mean much

[JazzLad]

And if there is anything we've learned from VW lately, it's that we can trust auto manufacturers.

Re:I don't think it will mean much

[0123456]

A vehicle won't be 100% autonomous, and least not for the foreseeable future. Meaning, there's still a steering wheel and peddles for human interaction.

But the fanboys keep telling us that all the human drivers will be gone in five years because Google.

Why would I want a 'driverless car' if I can't sit in the back drinking whiskey because the car might expect me to take over at any second? What's the point?

Re:I don't think it will mean much

[drinkypoo]

"Meat stock, you're revving up a slippery slope. I'm overriding that shit."

Meat stock? That's only after you're in a severe crash, and all that's left of you is soup. Anyway, traction control is awesome. If you have some actual traction to work with, and your TC is four-wheel, then it is ridiculously great.

Re:I don't think it will mean much

[Aighearach]

There aren't any real mysterious hurdles, you just deposit n dollars into a bank account, as per State or Federal self-insurance guidelines, and the government stamps the insurance cert as soon as the bank verifies the account. Even for individuals, self-insurance doesn't require a bunch of hurdles, it requires depositing real money into an account, having the State verify the amount, and then in some States you have to show them that the money is still there every year or two.

But they won't, because they'd have to manage it, and that's not their competency. They already have lots of insurance, just to have employees and work sites, and this is low risk so they'll just package it in with their existing insurance.

Aircraft already have significant manufacturer liability insurance that is paid when the vehicle is made, so there are already formulas to calculate the price based on accident rates and vehicle life.

Re:I don't think it will mean much

[Aighearach]

Sit in the back seat. Or don't buy one.

Honestly, until they get the issues of liability sorted out, the self driving car is a complete non-starter .. precisely because of crap like this.

You're claiming that if they don't "sort out" the imaginary "issues" that the whole product is a "non-starter." I would like to point out that if they don't come up with a new system, I'll just put the self-driving car on my existing insurance, and can have all the other benefits of the product. The rates will be low as soon as there is vehicle crash data available, which will happen before they even go on sale. That data is already being collected and analyzed by actuaries!

Re:I don't think it will mean much

[Aighearach]

Volvo isn't making the expansive claim the Google is. The gesture doesn't mean much, as you said. You're correct, your insurance will pay, and they may or may not then try to blame Volvo based on this. But the part you miss is that if they fail, or succeed, they won't be telling you about it. And it won't come back to you. It can't; they already paid your claim, and what happens with Volvo isn't a new fact about your accident.

Your guess about it not being cheap is silly, there is already public crash data on these cars. Insurers don't get squeamish about computers and AI, they just hire actuaries who use numbers to quantify risk. The risk will be low, and the insurance rates will appear low to customers used to normal auto insurance, but the insurance profit margin will skyrocket. Drivers and insurers will both be loving it.

You're right that Volvo is just re-stating a generic manufacturer's warranty, they're not actually promising to insure the vehicle the way that Google is. At least, based on the media statements they made about if the driver was "misusing" the vehicle. In the US, you're required to have liability insurance that covers that exact case, so I expect they'll actually walk back that part and provide real "liability insurance" and not just the thing they explained poorly using contradictory language. Presumably what they'll really offer is normal liability, with a limited comprehensive that has the disclaimer.

Re:I don't think it will mean much

[Aighearach]

Volvo is offering to indemnifying individual owners against flaws in the self-driving system. Of course, you'd have to prove somehow that the self-driving system was responsible, and do it by going up against a massive corporation's legal department.

On the other hand, if your self-driving car has a crash, who do you think people are more likely to sue? You, or the corp with deep pockets? Lawyers will probably be lining up for contingency fees to go after the corp.

You betray your ignorance of insurance. That is how it already works; you don't sue an insurance company, you sue the driver, and the insurance company provides their lawyer, and then pays the claim. Story time. My friend was a passenger in his girlfriend's car, and they were in an accident. It was ruled "no fault" (translation: both drivers made mistakes) and so his gf's insurance company was liable for his injuries. But they denied the claim, even though it was a rather obvious situation. In order for them to even negotiate the payment, he had to file a lawsuit; then of course they settled the claim. But who did he have to sue, the insurance company? No, he had to sue his girlfriend, and the lawyer had to write the complaint up using phraseology that blamed her for not paying his injuries, instead of the insurance company!

So it will be exactly the same as now. You sue the driver, or if the driver was not a legal entity, then you sue the owner of the machine. If they have insurance, the insurance will provide the lawyers and pay the judgement. Their liability is limited by a bunch of things, including the fact that they're not even the ones being sued, and also that the courts tend to view the insurance liability requirements as limits to what is normally recoverable. And indeed, most states cap it at that level.

Any big lawsuits will be related to recall type of situations, not accident liability. If a self-driving car kills your next of kin, you're only going to be able to sue for $25,000-50,000, depending on the State.

Re:I don't think it will mean much

[Altrag]

Volvo (the driver) is liable

There's trickiness to that statement. In particular, keeping up proper maintenance on your car is fairly important in order to keep things running optimally.

Presumably if Volvo (or Google or whoever) is willing to take on this liability, they're probably fairly confident in their sensors being able to detect anything too far out of whack and prevent the car from moving until its repaired (or at least shut off the auto-driving software along with a clause that their liability only covers times that its on.. which will likely be the case anyway because dumb people are dumb even if they keep their car in tip top condition.)

But it still leaves the question a little bit open.. driver claims he did everything right and it was the sensor that failed vs Volvo claiming the guy's an idiot and their car worked fine and so forth. Its not quite as cut-and-dry as it appears on the surface, and of course Volvo will almost certainly be looking for any possible loophole to get themselves out of any liability that could theoretically arise.

Re:I don't think it will mean much

[Altrag]

being required to take over in the event of an emergency

Emergencies are when the cars are most likely to NOT want the meatsock taking over. People are really really bad at making decisions while panicked. Pretty much anything the software can do to minimize impact in the case of emergency is likely going to be a smarter idea than what 95% of the population would do, even if they were in active control and alert the entire time.

The type of things that would likely be passed to manual control would be "we've arrived at the mall parking lot and I'm not going to even attempt navigating that shit." Times when its not likely to make a lot of difference whether the real person takes 5-10 seconds to get their shit together and take control.

Re:I don't think it will mean much

[Impy the Impiuos Imp]

It is conceivable they can be safer than human drivers, yet still get sued into onlivion by greedy lawyers who smell deep pockets, collapsing the industry and leaving road deaths greater than with.

Re:I don't think it will mean much

[david_thornley]

This isn't a big problem. You'll have to have your own insurance on the car for the foreseeable future (at least until the liability issues are sorted out), and so you're covered. Volvo's assurance might make your insurance less expensive (and just having a self-driving car might lower the rates also).

Re:I don't think it will mean much

[david_thornley]

It seems to me that $50K would buy a lot of insurance, and that's a significant sum of money to sit around in a savings account accruing approximately no interest.

Re:I don't think it will mean much

[BasilBrush]

But the fanboys keep telling us that all the human drivers will be gone in five years because Google.

No they don't. That's shorter than the normal life of the cars already on the road.

They may well say that autonomous cars will be in consumers hands within 5 years. And they may well be right.

As to whether the fully autonomous approach or the gradually add more automation to existing cars approach is better, we'll find out. American companies are generally trying the former. European companies the latter. Results, not theoretical arguments will provide the winner.

Why would I want a 'driverless car' if I can't sit in the back drinking whiskey because the car might expect me to take over at any second? What's the point?

What's the point of automatic transmission? Cruise control? Automatic braking before collision? They are all steps on the way to partially automated cars. And they are increasingly popular. And at the top end automatic lane following is already available to the consumer.

Re:I don't think it will mean much

[dj245]

It seems to me that $50K would buy a lot of insurance, and that's a significant sum of money to sit around in a savings account accruing approximately no interest.

The GP specifically mentioned a surety bond on file with the DMV. A surety bond is not a cash bond. In fact, it somewhat resembles insurance.

Surety bonds basically work like this- I go to my bank and say I want a surety bond for $10,000. The bank checks my financial situation and says "ok, DJ245 has enough assets, he could easilly come up with $10,000 if he needed to". I pay my bank a small fee (around 1-2% of the bond amount) and they basically write a letter that says "DJ245 can come up with $10,000 if he has to. If he can't, then we will pay it on his behalf and then we (the bank) will be responsible for making him pay us".

The fee to make the surety bond is nonrefundable. It is not the same thing as having a bunch of money in an account accruing no interest. A surety bond is a promise that payment will occur no matter what happens. That promise is based on verifying assets of the 1st party, and a guarantee that a 3rd party will pay if the 1st party can not.

Re:I don't think it will mean much

[mjwx]

Volvo is offering to indemnifying individual owners against flaws in the self-driving system. Of course, you'd have to prove somehow that the self-driving system was responsible, and do it by going up against a massive corporation's legal department.

This.

The first reaction of any warranty or liability claim is to find a way to blame the user. The most infamous of these in recent memory is Apple's Antennagate. More related to the Auto industry, look at how long GM ignored the ignition issue, it had to kill a dozen people before GM stopped denying it was even a problem.

Any acceptance of liability is going to be conditional, even though I'd trust Volvo far more than any other car company when it comes to safety (an old 240 is safer than some new cars, the B pillars are so strong you can rally in those things without fitting a roll cage, I'd still fit the roll cage though) this is still going to come with an 800 page document of legalese around the term "accept".

Re:I don't think it will mean much

[Joe_Dragon]

if it does not stop the car some where unsafe and waits / vs go into a limp mode to get some where safe to stop.

Re:I don't think it will mean much

[jabuzz]

Seems to me that a $50k insurance bond is utterly inadequate. It is perfectly possible for an vehicle accident to cause substantially more than $50k worth of damage. For starters medical bills are likely to be more than that, and even here in the UK the NHS is supposed to recover the cost from the insurers. Even putting aside medical bills that can easily run to seven figures, plenty of cars cost more than $50k so you so something stupid cause a crash that results in someone else's $100k car being written off? How about you loose control leave the road and demolish part of a house causing $100k worth of repairs?

Re:I don't think it will mean much

[Chatterton]

If volvo will try to not pay, then your own personal insurer would have to pay. And because they would not like to pay for Volvo, they will go against Them. 2 behemoth fighting against each other, could be fun to see, but you could wait some time before getting your indemnity except if your insurer make you an advance payment waiting for the mess to sort out...

Re:I don't think it will mean much

[delt0r]

The issues of liability for machine failure have been worked out. ABS breaks, just plain old breaks, tire blowouts, etc. Failure of systems that can cause potentially fatal accidents is *nothing* new. Why everyone thinks it is, is beyond me.

Re:I don't think it will mean much

[david_thornley]

Regrettably, in the US a $50K limit is often considered sufficient. In my state (Minnesota) the legal minimum is $30K per injured person and $10K for damage to property, with a cap of $60K. I believe that if the costs to pay out exceed that you sue the person directly, if said person has enough assets to make it worthwhile. There are reasons (like the ones you mention above) why I carry far more insurance than that.

Re:I don't think it will mean much

[david_thornley]

Thank you - missed that.

All car companies will have to do this

[gurps_npc]

Eventually they will see it as a "feature", rather than a bug. Buy our car and WE pay for the insurance. Of course, in reality, the price of insurance will be bundled into the vehicle.

Also, self-insuring is not as big a deal some people seem to think it is. Yes, there will be some legal/regulatory hurdles, but a lot of the that has to do with financial resources to pay it off, which VW will either still have or be out of business.

More importantly, it will eventually lead to huge profits as current computers are already far safer drivers than human beings.

Always remember it's like being chased by a bear - you don't have to be faster than the bear, just faster than your competitors.

Re:All car companies will have to do this

[bsolar]

More importantly, it will eventually lead to huge profits as current computers are already far safer drivers than human beings.

Not necessarily since a lower risk should translate into a lower insurance premium. Actually in some fields it's very strictly regulated and the insurance company is mandated by law to pay back to the insured any risk-based surplus within a few years. Of course if you instead give the insurance free rein...

Hmm

[koan]

I'm curious about the point in which we have autonomous and "manned" cars on the road simultaneously.
What sort of response will the human make to machine organized traffic.
Road rage maybe? Or a more peaceful driving experience, until that is, autonomous is all there is.

Re:Hmm

[Obfuscant]

What sort of response will the human make to machine organized traffic.

A more important question is how autonomous vehicles will respond to human drivers, and also how will they respond to a mass of other autonomous vehicles?

Emergent behaviour in simple automated systems is a wonderful field of study, and often results in postings in Risks Digest. It isn't always so easy to predict what a mass of simply programmed things will do, which is why it is called "emergent" behaviour and not "easily predictable".

Or a more peaceful driving experience,

Autonomous vehicles will be perfect machines, obeying traffic laws and having instantaneous reaction times. That means they will obey every law, even ones that are ridiculous and human drivers have learned to skirt to improve the efficiency of the system. For example, in a college town, traffic will come to a complete stop at class change time as pedestrians fill the crosswalks. Autonomous vehicles will obey the law that says they must stop when a pedestrian is "in the crosswalk"; most human drivers will take advantage of the gaps when one pedestrian has just left the crosswalk and the next one is on the other side of the road but not in danger.

I do predict that bike riders will be taught that THEY better obey the law, because autonomous vehicles will expect them, as vehicles, to do so, and if bike blows through a stop sign he's going to get his by the autonomous vehicle who is obeying the law and has the right of way. That, I say, is a good thing.

They will also behave in ways that take advantage of their reaction time, so that what humans would consider tailgating because their own reaction times would not allow it to be done safely autonomous vehicles will have no reason to avoid.

Re:Hmm

[cfalcon]

Your post calls it to mind, so I'll bring up a side rant I have about self driving anything, that never seems to have a place to go.

Pretend we, after massive investment and development, have self driving cars that work well and work almost everywhere.

Suddenly, Professor Light or whoever comes up with a new type of vehicle. What this is exactly doesn't matter, but the self driving cars don't understand it at all- they think it's a pedestrian when they should think it's a vehicle, and it's a technical challenge.

No matter how great Professor Light's MegaVehicle X is, now it won't sell. It's now up to him to somehow solve the problem for all the self driving cars- and of course, they aren't in the market for the new vehicle, and they are established as fuck, so the new better vehicle simply never is allowed on the roads.

If this seems a bit too sketchy, pretend that no one had invented motorcycles or bicycles or semis, and that all the development made assumptions based around that. The moment you try to add those to a real street with fully human drivers, it's fine- but do that with self driving cars and you'll be begging congress for a trial in Nevada and trying to convince investors that you can make the four competing driving systems recognize what a motorcycle is, and that the market will be people who don't want to own a full car. And then you have to get people who make cars for a living to support that.

The self driving car thing will paralyze the ability to innovate a goddamned thing in the transportation sector.

Re:Hmm

[lgw]

Google's self-driving cars are already being programmed to cheat a little on traffic laws - there was a /. story about that recently. I think autonomous cars will end up forcing changes in traffic laws in bigger ways, perhaps the first real re-think since cars became common. For example, once self-driving cars dominate, why have a government-imposed speed limit? The software will eventually become quite accurate in picking the max safe speed given 100 independent variables.

But bicyclists obeying the rules? Now you're off in fantasy land.

Re:Hmm

[Obfuscant]

It's now up to him to somehow solve the problem for all the self driving cars- and of course, they aren't in the market for the new vehicle, and they are established as fuck, so the new better vehicle simply never is allowed on the roads.

Software updates for autonomous vehicles will be possible from day one of them being released to the general public. It will be a simple matter for manufacturers to release upgrades that recognize the new kind of vehicle and make decisions that are right 100% of the time about those new vehicles.

Unfortunately, the update process will be through a wireless system using either the cellular data network or an adaptive mesh of short-range vehicle wireless data systems, or a combination of both.

And there will be no possibility of finding any means of hacking this global data system, just like it is impossible to control any functions of certain vehicles today using wireless data connections. The system will be perfectly safe, making perfect decisions perfectly always.

Your Friend's Job

[Anonymous Coward]

How long will your friend have a job if insurance companies only have to deal with a few car companies?

How long will car insurance companies be around? The car manufacturers will self insure with re-insurance to stave of massive catastrophe.

Claims adjusters are pencil pushing paper shufflers. I'm married to one. Assuming self driving cars and the inevitability of the manufacturers matching Volvo's tactic, there will be no claims adjusters.

Re:Your Friend's Job

[Hotawa Hawk-eye]

As long as you have a non-self-driving car, you're probably going to want or be required by law to have insurance.

As long as your self-driving car has a manual override and the car company's insurance guarantee won't cover incidents when that override is engaged, you're probably going to want or be required by law to have insurance.

Eventually, will there be no need for automotive claims adjusters? Perhaps. How long will it be before "eventually" occurs? A long time IMO.

Re:Your Friend's Job

[TWX]

Auto insurance will be around as long as there are uninsured drivers and other incidents beyond the fault of the owner or authorized driver.

Rates may reduce dramatically if liability is reduced to those cases, but the need for the insurance will still be there.

Re:Your Friend's Job

[Firethorn]

How long will your friend have a job if insurance companies only have to deal with a few car companies?

Given that he's a claims adjuster and not a salesman, his job should be fairly secure - he might have to scale back from 40 hours to 30 hours.

Remember, he's not just adjusting claims for on road accidents, but things like windshield repairs, vandalism, theft, etc...

For that matter, you might end up with an interesting split - liability is taken by Volvo for any damage caused by the car, including something like hitting a tree. It's still a good idea for the driver to pick up:
Under/Uninsured Motorist, theft, non-moving accident, etc... IE the rest of the 'comprehensive' package.

Then again, you might end up with Volvo being a bit like GM once was - before it's bankruptcy GM had essentially turned into a loan company than happened to sell cars.

BTW, car makers/car dealers offering insurance coverage on the sale of new cars is not a new thing over in Europe. It's quite common for a young adult's first car to be a brand new econo-shitbox because the provided coverage makes it cheaper than anything else.

Re:Your Friend's Job

[mjr167]

Is that a bad thing? Life moves on... when your industry becomes obsolete, you have to move on too.

This guy should be a lawyer

[JustNiz]

>> Of course, this is limited to flaws in the self-driving system.

Oh your car chose to kill a kid on a bike instead of hit an old person crossing the road? Yeah sorry you're on your own since we arbitrarily choose to not identify that as a flaw in our system.

Re:This guy should be a lawyer

[VorpalRodent]

You've just hit on an interesting scenario that will be to Volvo's advantage.

Volvo is driving. For any accident, they accept full responsibility. However, a holy-crap scenario arises where the computer has no viable options. Clearly, Volvo is still fully accepting responsibility.

Except, in that type of scenario, I'm going to grab the wheel and try to do something. Since I've done something in this worst case scenario, their lawyers will cite the computer data indicating that 1.4 seconds before the accident, the human driver took control. Suddenly I'm at fault.

Re:This guy should be a lawyer

[khasim]

Oh your car chose to kill a kid on a bike instead of hit an old person crossing the road?

What? Why do you think that a car would be programmed to hit "obstacle B" when "obstacle A" appears in front of it?

Instead, wouldn't the car be programmed to avoid ALL obstacles and apply the brakes with maximum efficiency?

Re:This guy should be a lawyer

[The-Ixian]

Exactly this.

As a matter of fact, the computer will know about the problem long (hundreds of milliseconds) before you see it and will already be reacting.

The idea that you could react faster or make a better critical decision than the computer is sort of funny actually.

Re:This guy should be a lawyer

[0123456]

Instead, wouldn't the car be programmed to avoid ALL obstacles and apply the brakes with maximum efficiency?

Yes, because cars can always stop instantly when a kid runs out in front of them.

Ultimately, someone's going end up programming the 'what to do when you have a choice between hitting a kid or a bus full of nuns' case, either intentionally, or as a consequence of general obstacle programming.

Re:This guy should be a lawyer

[0123456]

The idea that you could react faster or make a better critical decision than the computer is sort of funny actually.

Belief in the infallibility of computers and programmers is sort of funny, actually.

I test-drove an SUV last year which would beep if you started crossing the lines in the road. Wow, brilliant, right? Except, for at least six months a year, you can't see the lines on the road around here.

Re:This guy should be a lawyer

[SuiteSisterMary]

I think he refers to an old philosophical question.
The classic example is 'you're conducting a train. You come around a bend, and there's a track split. One track A is, say, a person. On track B is, say, two people. You don't have time to brake. All you can do is pick which track you take. Which one do you take?

What if Track B has five people? One child? A world-class doctor who saves lives? A scumbag criminal? Your wife?

So, say you're in a self-driving car. The car wants to make a left turn across traffic at a four-way intersection. So it advances into the intersection, stops to wait for a break in oncoming traffic, and waits.

Sensors notice a semi coming up behind you, and not stopping. In front of you is an old person crossing the road. To your right is a kid on a bicycle. To your left is a stead stream of through traffic. Where do you go?

Re:This guy should be a lawyer

[0123456]

No. In that scenario, they'll presumably do what aircraft manufacturers do. The autodriver will turn off, and they'll blame 'driver error!' when you crash.

Re:This guy should be a lawyer

[elgatozorbas]

The idea that you could react faster or make a better critical decision than the computer is sort of funny actually.

Your reaction to this article is also sort funny. The situation you describe, with the computer making faster and better decisions, is what you expect to happen. A desirable output. And most of the time this is what WILL happen.

However, in liability issues we are not interested in the desired result. We are interested in failures. When things go wrong in whatever imaginable or unimaginable way. As this Volvo CTO seems to be more aware of than you, this WILL happen some day. That will most likely be the same day that the GP's prediction comes true, and Volco announces it doesn't count as a flaw in their system.

Re:This guy should be a lawyer

[Obfuscant]

As a matter of fact, the computer will know about the problem long (hundreds of milliseconds) before you see it and will already be reacting.

Knowing about a problem hundreds of milliseconds before a human does doesn't mean the computer will have a better solution than the human would. At some point, there will be the problem of a person stepping into a crosswalk inappropriately at the same time a bike rider blows through a stop sign and becomes an obstacle. When the solution is "slam on the brakes" because there are other people where the car would choose to swerve, and there is no physical way of stopping before hitting one of the new obstacles, the car will have to choose, and the car will kill someone.

At that point, there will be one of two results (besides someone getting run over, which won't be avoidable). First, the human in charge will probably try to do something about the problem, even if the outcome is clear. Foot on brake, grab whatever wheel there might be, etc. The driver has interfered with the car's operation and is therefore accepting full liability for the result, and the car maker is off the hook.

Or second, the driver will hear an odd sounding "bong" noise, a small red icon on the dash will light up, and the car will have cancelled "autopilot" mode. (That's what happens when autopilots in aircraft turn themselves off. In an airplane it's called an annunciator and will probably be displaying "autopilot disengaged", but you have to know where to look for it.) At that point, the car is no longer driving, so the car company is off the hook again.

The idea that you could react faster or make a better critical decision than the computer is sort of funny actually.

The idea that computers can be programmed to include every possibility that a human might be able to imagine to solve a problem, and have the physical ability to carry any necessary preventative action to completion prior to an unfortunate result, would be sort of funny, were it not the baseline assumption of so many people pushing this technology. Perhaps it's just because I've read Risks Digest for so many years and deal with these perfect computers on a daily basis that I don't assume the robot overlords will be better for us?

Re:This guy should be a lawyer

[laie_techie]

Belief in the infallibility of computers and programmers is sort of funny, actually.

I test-drove an SUV last year which would beep if you started crossing the lines in the road. Wow, brilliant, right? Except, for at least six months a year, you can't see the lines on the road around here.

Not to mention the other 6 months of the year when there's road construction and multiple markings on the road.

Re:This guy should be a lawyer

[mayko]

Exactly this.

As a matter of fact, the computer will know about the problem long (hundreds of milliseconds) before you see it and will already be reacting.

The idea that you could react faster or make a better critical decision than the computer is sort of funny actually.

I agree that a computer will certainly be able to react faster in the event of a sudden unexpected obstacle, but what about my human ability to see children playing near the road or someone who isn't looking my direction but hasn't actually stepped off the curb? I can preemptively slow my vehicle just-in-case. A human can quickly detect things about their surroundings that computers aren't yet capable of.

Re:This guy should be a lawyer

[Obfuscant]

The car wants to make a left turn across traffic at a four-way intersection. So it advances into the intersection,

The car has just broken the traffic law. It has entered an intersection prior to having an ability to complete the action and leave the intersection. In some places this is made explicit (large cities, e.g.) by painting a box in the intersection and actually ticketing people who cause "gridlock".

Your autonomous vehicle will sit patiently at the stop line until traffic clears enough to be able to complete the left turn, even if that means it never completes the left turn. A human driver would do as you say because it is more efficient and carries little risk to anyone.

Sensors notice a semi coming up behind you, and not stopping. In front of you is an old person crossing the road. To your right is a kid on a bicycle. To your left is a stead stream of through traffic. Where do you go?

This dilemma is valid in any case. The only course of action for the autonomous car is to do nothing. A human might trust that a pedestrian will also notice the problem and jump out of the way, but an autonomous vehicle cannot predict such behavior. You're toast.

Re:This guy should be a lawyer

[Ichijo]

Belief in the infallibility of computers and programmers is sort of funny, actually.

The Space Shuttle Columbia's computer did a remarkably good job of keeping the nose pointed in the right direction while the craft was disintegrating, right up until the time it lost hydraulic pressure and control became impossible:

the [plasma] breach ultimately caused unusual aerodynamic drag to develop on the left side of the spacecraft, forcing Columbia's flight computers to adjust the shuttle's roll trim with the elevons, or wing flaps, on each wing. Eventually, two right-firing rocket thrusters were ignited to provide additional muscle. But it was a losing battle.

Re:This guy should be a lawyer

[0100010001010011]

Version 1 of something didn't work guys. Lets pack up all R&D and give up.

Re:This guy should be a lawyer

[adiposity]

What if "obstacle B" is a cardboard cutout of Mickey Mouse and "obstacle A" is a person? Maybe you actually could make a better decision sometimes.

Actually, the idea that computers always will make a better decision than a person is hilariously optimistic. Sure, they will make a faster decision, which most of the time will avoid both obstacles, but that's not always the same as better.

There's no doubt in my mind that self-driving cars will improve traffic immeasurably, because of quicker reaction times, but computer-based systems frequently do really stupid things that no human would do. Putting one in a car doesn't make them flawless.

Re:This guy should be a lawyer

[Aighearach]

Oh your car chose to kill a kid on a bike instead of hit an old person crossing the road?

What? Why do you think that a car would be programmed to hit "obstacle B" when "obstacle A" appears in front of it?

Instead, wouldn't the car be programmed to avoid ALL obstacles and apply the brakes with maximum efficiency?

They're confused into thinking this way because their driving practices are so dangerous, they actually plan ahead to swerve around an obstacle instead of stopping before hitting it. These are exactly the idiots who will stop running over Jr and Grandma when they switch to a self-driving car that will follow the DMV mandate and stay in the same lane and stop before hitting anything.

I kinda think the traffic engineers are right about this one. If you had time to "choose" what to hit, you'd be driving a safe speed and could choose not to hit anything. Actually, self-driving cars excel at stopping quickly when grandma wanders into the road. That has already happened a bunch of times in road testing. They're not going to be distracted, which is the main danger.

Re:This guy should be a lawyer

[Aighearach]

You erect the straw man "computers are infallible" to attempt to defeat the claim that computers react more quickly than humans. Fail.

Also, the thing you drove wasn't a commercially available self-driving car, it was a different thing, very primitive with a limited intended function that is different than a self-driving car. One could almost think you were comparing apples to oranges, but in this case it is more like comparing an apple to a cartoon orange sticker.

Re:This guy should be a lawyer

[JustNiz]

So you seriously think there could never be a situation where the world didn't behave as planned and it would have to make a choice between "Evil A" and "Evil B"?

Re:This guy should be a lawyer

[JustNiz]

I may not be able to (re)act faster but I think real world situations still exist where I could make a better decision than a chip.

Re:This guy should be a lawyer

[Aighearach]

Because of the way insurance works, that won't be a real issue. What they offer will either meet the requirements of auto liability insurance, in which case it won't matter what their analysis of the fault is, or it won't meet that requirement, and you'll have additional insurance that actually covers the accident.

Accident fault is based on traffic laws and which vehicle went where at which time. If they own up to a technical fault will be a separate issue for them to fight out with the insurance companies. It won't come up during the auto liability cases, which are heavily constrained by existing law and precedent. It will be the same as now; if your car has a manufacturer fault causing your wheel to fall off and hit another car, then the accident was your fault and your insurance will be covering it. If you want to blame the manufacturer, or your insurance wants to, that will be a separate case where having to pay the cost of the damage to the other vehicle is the harm you're suing them for. As it is now, "mechanical fault" officially accounts for over 10% of vehicle accidents. So this is not even close to new legal territory regarding the liability.

Re:This guy should be a lawyer

[Aighearach]

The part you're getting wrong is that it is never appropriate to swerve. If you had time to check that it is safe to do so, you'd have time to stop. If somebody steps into the crosswalk when you don't have time to stop, your duty is to brake as quickly as possible to reduce the speed before impact. DO NOT SWERVE. The self-driving car is going to get this right 100% of the time; it won't be programmed to panic and create a new accident because of a rash action. Those milliseconds will directly translate into quicker stopping time, or at least reduced impact speed.

Re:This guy should be a lawyer

[Aighearach]

That is pretty easy. You don't "go" anywhere, you make sure your arms and face are in a position to minimize injury when the airbag deploys, if the semi in fact doesn't stop in time. Duh.

And as a moral thought exercise, you don't address the issues raised in the train situation. That exercise is based on the fact that you can't stop the train; there is no neutral course of action. In your situation, you do have a neutral course of action that doesn't involve killing anybody, and so there is no actual moral exercise. It will never be morally acceptable to risk killing somebody in order to reduce a risk to yourself that is uncertain and outside your control. There is no way that you have time in that scenario to weigh the risks, so you can't reach the position of claiming that a balance of risk favors an action. Therefore, you can't justify actively harming any of these people; even if you are OK with harming an innocent party to reduce your own risk when the risk to you is greater than to them. So the only moral choice you have is the same as what the DMV says you're legally required to do; stay in your lane, and remain stopped until the danger has passed.

Like all the other situations people propose where one of the considered actions is to leave your lane and intentionally crash into something; the answer is "don't." A self-driving car is going to get this right 100% of the time. If you absolutely are going to be in an accident, and there is no way to avoid it; continue following the traffic rules. They were designed by traffic engineers with this in mind. Stay in your lane, and stop. Easy.

Re:This guy should be a lawyer

[bsolar]

Because brakes might not be enough: then the decision becomes: "do I hit the obstacle or attempt to dodge it? Hitting the obstacle might mean a lower risk for the passenger of the car, but a huge risk for the obstacle. Attempting to dodge might mean a higher risk for the passenger of the car but a much lower risk for the obstacle.

If the obstacle is a dog you might want to prioritise the passenger of the car's safety, but if the obstacle is a kid you might want to attempt to dodge even if the passenger is put at higher risk.

Re:This guy should be a lawyer

[Obfuscant]

The part you're getting wrong is that it is never appropriate to swerve.

Tell that to the mother of the child you've just run over because they stepped out from between parked cars and you didn't swerve over into the other lane, you just slammed on the brakes and ran them down.

Tell the mother of the child you just ran over that you didn't attempt to swerve because you might have dented the fender of the car in the next lane over.

If you had time to check that it is safe to do so, you'd have time to stop.

I'm sorry, but that's patently absurd. As a defensive driver, I see and keep track of vehicles coming my way in the other lane. I can't see a four year old child running into the street from between parked cars until they do it. It takes zero time to judge that it is safer for that child if I enter the oncoming empty traffic lane than if I attempt to continue straight ahead.

Now, I won't ALWAYS have time to determine that it is safer to change direction. I might be in a multi-lane highway where someone is approaching from the next lane over. There might be a curve I can't see around. But you said NEVER, so all I need to do to disprove your statement is show that it is possible.

If somebody steps into the crosswalk when you don't have time to stop, your duty is to brake as quickly as possible to reduce the speed before impact.

You're saying it is my DUTY to run them over instead of take evasive action? Wow. I hope you have nothing at all to do with programming these new wonderful autonomous vehicles.

Re:This guy should be a lawyer

[Obfuscant]

They're confused into thinking this way because their driving practices are so dangerous, they actually plan ahead to swerve around an obstacle instead of stopping before hitting it.

Defensive driving requires thinking ahead of time what one would do when presented with a surprise on the road, such as a child popping out from between parked cars. You cannot say "I plan on stopping before hitting a child that does that" because you cannot plan on it happening far enough ahead of you that you could stop. The best drivers will ALWAYS think ahead far enough that they know "should I be unable to stop from hitting a child in the street, there is an open space I can use to avoid it."

It's called "giving yourself an out", and sometimes it does, indeed, mean you'd swerve into an empty lane to go around a sudden obstacle.

The dangerous ones are those who make statements like "autonomous vehicles will get it right 100% of the time."

a self-driving car that will follow the DMV mandate and stay in the same lane and stop before hitting anything.

Come drive in the real world and you might learn that DMV cannot mandate that you stop before hitting something that appears in your path without notice. DMV might be all powerful on your planet, but on Earth they are still limited by the laws of physics. In the real world, you will find that obstacles sometimes pop out of the underbrush on the side of the road and run straight into your path much faster than you can even see them, much less come to a complete stop to avoid them. I've had deer, possum, and even a skunk that decided that they want to cross the road RIGHT NOW and didn't care that there was something with big bright lights coming at them. For all except the deer, fortunately, I didn't have time to even really see what it was before it disappeared under the front of my car. The deer was fast enough to get across before being hit.

If you want to argue that simply hitting such an obstacle is the right answer, well, consider that small children can also decide they need to cross the road RIGHT NOW and often don't care that something big is already using that part of the road, or is about to. And they may not hide in the weeds along the road, they'll hide in between parked cars or behind other things.

Now, you might try to argue that traveling at 10MPH on a major US highway is the safest way to prevent that, but I think most people would disagree. Again, I pray that you have nothing to do with programming autonomous vehicles because you would have them all going at 10MPH on every road just in case they needed to stop suddenly.

Re:This guy should be a lawyer

[Obfuscant]

So you seriously think there could never be a situation where the world didn't behave as planned and it would have to make a choice between "Evil A" and "Evil B"?

Be calm, citizen. The autonomous vehicle will make all such choices for us, and the autonomous vehicle will get the answer right 100% of the time. You never again need to worry about what was right or prudent or safe, the decision will be made for you. People who think about this all day, every day, for a year will design and accurately code the correct responses to all such possibilities, and you cannot possibly know a better answer as a non-engineer.

Please enjoy your trip in comfort and forget about the outside world for the duration.

I was going to add that if you ever pondered or debated whether "Evil A" or "Evil B" was the better choice, you could simply set up Evil A, Evil B, and an autonomous vehicle and let it make the decision. Whichever it chooses is obviously right, 100% of the time.

Re:This guy should be a lawyer

[Obfuscant]

If you absolutely are going to be in an accident, and there is no way to avoid it; continue following the traffic rules. They were designed by traffic engineers with this in mind. Stay in your lane, and stop. Easy.

A child has jumped into the street without warning from between parked cars. I could move over into the empty lane beside me, but there is a double yellow no-passing stripe. I must follow the traffic rules and hit the child instead of breaking the law that says I cannot cross the double yellow.

Of course I must then stop, otherwise I will be committing the felony of hit-and-run, and I must obey all traffic rules -- even if the death of a child is the result. I can't say "an innocent child" because the child is the cause of its own demise and is not innocent as the result. I am, however, innocent, because Aighearach said so.

Re:This guy should be a lawyer

[amiga3D]

He didn't say computers were infallible, just better than the average driver. That's not hard at all.

Re:This guy should be a lawyer

[mjwx]

Oh your car chose to kill a kid on a bike instead of hit an old person crossing the road?

What? Why do you think that a car would be programmed to hit "obstacle B" when "obstacle A" appears in front of it?

Instead, wouldn't the car be programmed to avoid ALL obstacles and apply the brakes with maximum efficiency?

You make it sound like all collision are avoidable.

But arguing over which object it will choose to run into is pointless. We already know because we have these things called "road rules" and "safety procedures" already for meat based drivers. The fact that most meat based drivers have no clue what they are is not withstanding.

If someone cuts off an automated car and emergency brakes, yes it will run up the arse of it because that is the safest procedure (nose-tail accidents have the least severe injuries). No point in arguing that there is a Nobel laureate one car and someone who is actually important in the other. The car doesn't know and the car doesn't care because that has no factor on minimising injury.

Re:This guy should be a lawyer

[Aighearach]

No, your duty is to drive a safe speed and pay attention. And yes, to continue following the safety rules. What you don't seem to get is that in reality swerving kills somebody that should have lived, and doesn't save anybody because stopping is more effective. It isn't "save a child" vs "dent a fender," it is "do what you're supposed to do that is known to reduce fatalities" vs "do something you don't have time to measure the effects of, in a situation where you don't know what is going to happen." And no, if you could "see and keep track of vehicles coming my way in the other lane" you would have "seen and kept track of" the child and slowed down before creating this totally absurd false choice. Any time that you have time to "swerve" safely, you would have time instead to make a legal lane change if that will solve the problem, or simply stop if not. You're not allowed to drive faster than you can react to things in your lane, and in your scenario you're potentially killing a child by driving too fast when there is not enough lane clearance.

What will happen is, you'll swerve and hit a vehicle, cause a multiple car crash, kill a whole family, and the child will have jumped back out of the way anyways. So you'll have killed a whole family for nothing, just because you can't comprehend that traffic engineers are correct when they say NEVER SWERVE, STAY IN YOUR LANE AND STOP

Re:This guy should be a lawyer

[Aighearach]

LOL, you're pretty deep into justifying awful driving when you're trying to convince yourself, and others, that swerving counts as defensive driving.

Re:This guy should be a lawyer

[Aighearach]

Odd that you go so crazy in trying to create a situation where it is morally justified to swerve you had to presume that the person swerving is such an immoral asshole that if he hits a pedestrian who dove in front of his vehicle, he'll just drive off.

If you had time to swerve, you had time to stop. There are not rows of parked cars on the freeway. Streets that have lane-side parking have speed limits such that if you have time to turn the wheel to swerve, you'd have time to stop too. And if children are just popping randomly out from behind parked cars, and you can't see that kids are playing by the road as you approach, how the hell are you going to know if a bicycle just pulled out into the other lane and you didn't notice yet?

There is a lot more going on in this scenario than just an unseen kid. There is a speeding asshole who thinks he's smarter than an engineer, and a child whose parents' fault the accident would be, and the bicyclist who you killed who was the only innocent party in the scenario.

Re:This guy should be a lawyer

[strikethree]

Exactly not this. Perhaps in the situation that a child suddenly darts out in the road from behind an obstacle, the computer will react faster and better; however,the human brain is a VASTLY superior pattern matcher and can recognize that there might be a problem long before a computer even begins processing what to do about the problem.

Example: I am driving through a neighborhood moving at 20 mph. There was a child, perhaps 3, on a bike riding on the sidewalk. I saw that the child was unsteady so I slowed down even further, to perhaps 12 mph. Sure enough, as soon as I did that, the child lost control and dove straight in front of my car from the sidewalk.

A computer would have been going 25 because that was the speed limit. It would have reacted a few milliseconds faster than I did. I preplanned and only had to stop a multi-ton vehicle from 12 to 0. The few milliseconds that reaction time bought for the computer would be vastly overridden by the slower speed I was going. I saw a pattern that a computer would not.

To be fair, almost any other person would have ended up hitting the child too. I happen to have taught lots of children how to ride bikes and I could intuit what was about and what could happen. That child was super lucky that day.

Re:This guy should be a lawyer

[dave420]

Many straw men do not a compelling argument make. I'm sure you had fun concocting your illustrative fantasy, but it remains that - a fantasy. Thanks for playing.

Re:This guy should be a lawyer

[dave420]

You do realise that with every post like this you're broadcasting to the world what a terrible driver you are, right? Were you expecting people to read it and think "I, too, am a terrible driver. I now dislike autonomous cars because they can't screw up as much as I do"? The mind boggles...

Re:This guy should be a lawyer

[SuiteSisterMary]

Sorry, I'm talking Ontario traffic rules; when turning left on a non-advanced green, you advance into the intersection and wait for a break in oncoming traffic to complete the turn.

Re:This guy should be a lawyer

[strikethree]

I thought to myself: What is a person good at that a computer is terrible at?

Pattern recognition.

So I thought to myself, what incidents, while driving, have I been in where pattern recognition came in to play?

I thought of numerous bad weather scenarios and several vehicle ahead scenarios that I had been in but the one that struck me the hardest was the little boy on the bicycle. I felt truly justified with my thinking that day.

Do I care that you do not personally believe that it happened to me? Not really.

Do I have any idea what you mean by straw man? I know the meaning of the term but I have no idea what you are talking about.

Are you saying that kids do not ride bikes on the sidewalk? Are you saying that it is not possible to see that they are clearly VERY new at riding and could have troubles? Are you saying it is impossible to go from the sidewalk to the front right tire of a car in a moments notice? I can not quite figure out where the straw man is. I am weary of playing Where's Waldo. Enlighten me with an actual argument or just shut up.

Re:This guy should be a lawyer

[Obfuscant]

You do realise that with every post like this you're broadcasting to the world what a terrible driver you are, right?

Why yes, because deer, possum, and skunks, and four year old children, only decide to run into the roadway in front of terrible drivers, and they all know not to do so in front of such good drivers as yourself.

Don't be stupid. I drive in the real world where the unpredictable is unpredictable, and where a change of course is sometimes the best solution to a potentially lethal situation. I drive in a world where we value the life of the four year old who doesn't know not to run out into traffic from hidden locations over the arbitrary traffic rules of yellow lines painted on a road, and where only a heartless moron would maintain a straight-ahead course knowing he's going to run over someone's baby when there is an empty lane he could move into with no effort at all.

I don't know where you drive that it is "good driving" to go only 10 MPH on a major highway because you might need to stop for a sudden obstacle, but please don't do it on any highway on planet Earth. You'll be the hazard in that case, not the solution.

Re:This guy should be a lawyer

[Obfuscant]

Odd that you go so crazy in trying to create a situation where it is morally justified to swerve you had to presume that the person swerving is such an immoral asshole that if he hits a pedestrian who dove in front of his vehicle, he'll just drive off.

Actually, the person who DID NOT swerve is the immoral asshole who chose to run over an innocent child instead of swerve into an empty oncoming lane of traffic, and the one who did NOT swerve is the one who DID have the accident and must stop. He's the only one I talked about having to stop to maintain perfect adherence to the almighty traffic rules. That you didn't realize who I was talking about tells me you didn't read, or didn't comprehend, the hypothetical situation.

The moral, decent guy who chose to break a minor traffic law to save a life doesn't need to stop because there was no accident. The "hit and run" laws don't include "miss and run".

If you had time to swerve, you had time to stop.

From this table, I see that the total stopping distance for a car going 30MPH is 109 feet. That's ten car lengths. There are many streets in my town that have onstreet parking with a 30MPH speed limit. Even with 0 reaction time, the physical distance is still 43 feet. That's four car lengths, about. If you cannot make a full change of lanes in less than ten car lengths, you don't know how to drive.

Even at 20 MPH, the stopping distance is 63 feet. Ditto changing lanes in six car lengths. But you don't have to make a full lane change, all you need to do is avoid hitting a four year old child.

So, sometimes you may be able to stop before you hit the child. Sometimes you won't be able to, and a sane, rational, ethical human being will chose a path that will save the child rather than simply run it over because someone told them the only proper thing to do was "keep going straight ahead."

There are not rows of parked cars on the freeway.

There aren't. But there are often culverts, ditches, or grassy areas close. And those animals which decide to cross sometimes think if they go fast enough they'll make it -- faster than you can stop. I've had such animals dart into my path from less than 10 feet away, much less that 109 feet I'd need to stop for them going at just 30 MPH. The table says that even with zero reaction time involved, I could go no faster than 15 MPH and I'd still hit that animal. If you think driving on I5 or I90 or any other highway at 15MPH is the right thing to do so you will never have to avoid hitting an animal by anything less than coming to a full stop, you're dangerous.

Streets that have lane-side parking have speed limits such that if you have time to turn the wheel to swerve, you'd have time to stop too.

Nope. Maybe on your planet, but not on planet Earth. "Time to turn the wheel" is milliseconds" and is for the most part "reaction time". "Time to stop" includes "reaction time" plus the physical stopping action, which can be a lot longer.

And if children are just popping randomly out from behind parked cars, and you can't see that kids are playing by the road as you approach, how the hell are you going to know if a bicycle just pulled out into the other lane and you didn't notice yet?

I see kids playing by the road all the time. Do you really stop for each and every group of them, just in case? No, you don't. Neither do I. And I'm not counting them continuously, so if one of them goes in between the parked cars I may not notice that specific detail, until they pop out into traffic.

That empty oncoming lane, if there is a bike rider in it, will be obvious. He will most likely be on the other side of the lane to begin with, and I don't need the whole lane to swerve around a child just appearing from between the cars.

So your idea of safe driving is to take the known, sure

Re:This guy should be a lawyer

[Obfuscant]

No, your duty is to drive a safe speed and pay attention.

I'm doing that.

And yes, to continue following the safety rules. What you don't seem to get is that in reality swerving kills somebody that should have lived,

You speak in such absolutes. "Swerving kills somebody". No, it doesn't necessarily kill anyone. If I swerve over into an empty oncoming lane, I've killed NOBODY. There is no magic "action at a distance" where somebody dies just because I've broken a relatively minor traffic law.

and doesn't save anybody because stopping is more effective.

If you are physically able to stop in time, yes. If you aren't, then swerving is more effective. That's why a blanket "never swerve" is inappropriate.

It isn't "save a child" vs "dent a fender," it is "do what you're supposed to do that is known to reduce fatalities"

You to think that general probabilities rule the world, and then when faced with a specific situation that requires a specific decision to be made, you should ignore all the other factors and stick with the table of probabilities. Do you also refuse to use an umbrella while walking in a rain storm because the morning weather report said there was only a 20% chance of rain?

And no, if you could "see and keep track of vehicles coming my way in the other lane" you would have "seen and kept track of" the child

You have never driven in the real world. Your arguments are based on probabilities and wishful thinking about a perfect world. The world where computers will make the correct choice 100% of the time and have perfect ability to carry them out.

Any time that you have time to "swerve" safely, you would have time instead to make a legal lane change if that will solve the problem,

A legal lane change requires a relatively long period of signalling the intent ("failure to signal a lane change" is a traffic violation in most places, rarely ticketed on its own but a traffic rule nonetheless), and would be prohibited in many cases by local limitations such as "do not pass" or simply a solid white line separating two adjoining lanes. Swerving safely can be done "now"; changing lanes has to wait until you've obeyed the signalling requirement and may not be allowed by law even then.

You're not allowed to drive faster than you can react to things in your lane,

Uhh, what a strange planet you live on. On Earth there is no such law. There is "too fast for conditions" which involves consideration of the general condition of the roadway, but it would be impossible to go anywhere at all if the law required that you must not drive faster than you can react to the sudden appearance of an obstacle three feet in front of you without hitting it. And if you don't know that this can happen, then you don't have sufficient experience driving to be discussing it.

What will happen is, you'll swerve and hit a vehicle,

Not always, and never when I have done it. There simply was nothing there to hit other than the obstacle, and I didn't hit it. I would have had I followed your admonition to "NEVER SWERVE, STAY IN YOUR LANE AND STOP".

Your absolutism combined with lack of real world experience makes you dangerous.

Re:This guy should be a lawyer

[JustNiz]

>> you cannot possibly know a better answer as a non-engineer.

Ironically, I actually happen to be a software engineer in the auto industry. Looking at the weak software processes in the industry and the masses of crap code thats already deployed, there's no F'ing way I'd trust my life to a car. Hell from what I've seen in real life I wouldn't even trust it to navigate reliably.

After the 5th lawsuit ...

[fourthrail16309]

the car companys acceptance of "full liability" will end.

Re:After the 5th lawsuit ...

[bobbied]

No, the EULA on the software will be changed and THEN manufacturer liability will end when you accept the 30 page license after they push the next software upgrade Tuesday night.

Re:After the 5th lawsuit ...

[Joe_Dragon]

and then after you get out of your 5 year term in lock up you go and hunt down the ceo / cto / codes and beat the shit out them.

hmmm....

[CaTfiSh]

I hope it turns out better for them than when they held a press conference to demonstrate their new crash avoidance system: https://www.youtube.com/watch?... http://jalopnik.com/5533328/vo... ....or a few months later when their pedestrian avoidance tech proved to be a autonomous Death Race 2000 contender: http://jalopnik.com/5648126/vo...

Re:Good move

[Anonymous Coward]

What!? You mean I, as an armchair critic, didn't instantly and easily spot a fatal flaw that was overlooked by hundreds of experts on the subject matter who think more about the issue in one afternoon than I do in a year!?

Noooo! It cannot beeeeee!

And why not!

[DarkOx]

I can't imagine the big auto's pushing for any other policy. What better way to ensure that no new competition ever emerges. No more pesky start ups like Tesla showing up and disrupting things, nope if you are not already established with billions of dollars in assets, you need not apply.

Re:And why not!

[Aighearach]

So you're saying that new car makers won't be able to start up, and your argument is that Tesla drivers can't afford to buy their own insurance? Or that companies like Tesla, which exists only because a rich guy dropped a giant wad of billion dollar bills on it, won't be able to exist because you'll need billions to start a car company? What the hell, man, I know Thursday is the new Friday but you posted at 10:18am and it is way too early to be that drunk.

good move

[argStyopa]

This is definitely a context where Volvo's obsession with safety and their reputation for it will pay off in spades, both with regulators and the marketplace.

Re:Insurance Companies Are going to Go Batshit

[bobbied]

Not really... You will be required by state law to carry liability insurance which Volvo is not agreeing to provide. They are just agreeing up front to assume liability for accidents where the computer is driving, which is only part of what the state requires you to carry insurance for. Then there is the collision insurance that your lender/leaser will require that you carry as long as you don't own the vehicle outright. There will be plenty of insurance to sell...

Re:Insurance Companies Are going to Go Batshit

[Hotawa Hawk-eye]

They get to charge you 100% of the time for insurance and they'll only have to pay out in the small percentage of times that you have an accident after you take manual control over your vehicle. It's the politician plan: do nothing (or very little) and get paid. Sounds a bit like robbery to me.

Sure.

[c]

I think I'm going to need the see the fine print on that promise... the words "$corporation will accept liability" are rarely written without conditions a truck could drive itself through.

Re:Sure.

[Aighearach]

I think I'm going to need the see the fine print on that promise... the words "$corporation will accept liability" are rarely written without conditions a truck could drive itself through.

My advice is to not waste time on that. You'd be better served researching your State's liability insurance requirements. You might discover that if it meets the requirements of insurance, they have to pay and the Court doesn't care about their excuse. That's what liability insurance is; you screwed up, hurt or killed the other person, and now your insurance pays them. Saying it was really your fault because of fine print, that isn't a scenario that gets them off the hook.

OTOH if it doesn't meet the requirements of liability insurance, then you still don't need to care and can just learn about your insurance, because then it would be something that your insurance company would take up with Volvo after paying your claim, and you'd never even learn the result or be impacted by it.

Look at the fine print

[140Mandak262Jamuna]

You can't trust these European car makers. The fine print will say "contract valid only when driven on test tracks instrumented by Volvo a priori"

Re:Look at the fine print

[CanadianMacFan]

Yes because GM has shown a great social conscience when it comes to their customers' safety.

The side effect will hit regular drivers

[bjdevil66]

As more companies adopt this "we cover it" guarantee, they're going to start having more "costs" dealing with human drivers. Accidents, time/speed inefficiencies (vs. following the letter of the law), etc.

As a result, they'll start pushing the driverless cars harder. For those they can't convince with the carrot of incentives to make the switch, they'll eventually pull out the stick of passing the costs to those human drivers through various lobbying channels - forcing them to deal with higher insurance premiums, more tracking technology, tiered registration costs for cars, etc.

Corporations with any clout always try to cut their costs - one way or another.

Snow

[freak0fnature]

Given that self driving cars don't work in snow (let alone something as common and blinding as a good lake effect storm)...I don't see any effect on insurance at all, at least in the North East.

Re:Snow

[Aighearach]

Wags repeat this over and over, but in fairness, the vehicles had not yet been programmed for those conditions. It is the case that needs special work, and they didn't bother doing that part early on. So it tells us nothing about any of it to act like it is some sort of fact of nature that "self driving cars don't work in snow." That is a broad statement that may not even be true now. We don't know. They won't tell us that they're even seriously working on the problem unless they've already got it pretty polished. That is just how R&D works. The public can't see the progress as it happens. Everything on the road that is a "self-driving car" is a prototype. Many current ones may already have this fixed. I can tell you one thing, when they talk about timetables for wanting regulatory approval to actually sell them, they're taking into consideration that they will have to work in snow, and all other seasonally common weather conditions.

Re:Snow

[dotancohen]

Wags repeat this over and over, but in fairness, the vehicles had not yet been programmed for those conditions.

The issue as I see it is that if the self-driving car is used in the three good seasons, then in the snowy season _nobody_ will know how to drive. Not the car, and not the meatsack. The car has yet to be programmed, and the meatsack has no experience driving.

Re:Will the CTO be willing to due hard time it a c

[Moof123]

Corporations are people, except when it comes to criminal liability.

Car companies have a very poor track record when it comes to liability. They tend to fight responsibility for years. Is that what people are willing to accept if they have had their lives shattered by some badly written code that caused a car to cripple them?

Unless there is a clear and transparent translation as to what "liability" they are taking on means, we can expect that all lawsuits will get met with typical car company fervor, with stonewalling for access to technical documents and code being more likely than not. It is easier for a company with a fleet of lawyers to make these offers than for me to accept such an offer.

Re:Will the CTO be willing to due hard time it a c

[Obfuscant]

Corporations are people, except when it comes to criminal liability.

Right. Because you cannot put a corporation in jail, but corporate executives can and have been put in jail for corporate misdeeds. It is pretty easy to Google a list of high ranking CEO/etc corporate people who have gone to prison, or would have if they hadn't died before being convicted. Can we put to rest the canard that criminal liability is avoided by corporate "persons"?

"Corporations" have the right to free speech because the corporations are made up of people who do not abandon that right by forming a corporation. That's Citizen's United in a nutshell.

Car companies have a very poor track record when it comes to liability. They tend to fight responsibility for years.

That's absolutely true. Every company fights against liability suits. Just like most people who are sued fight back.

I see this as just another publicity stunt/spin tactic, and the result will be the answer "the driver is at fault for not taking control at the appropriate time", just as "pilot error" is the usual cause of aircraft accidents.

Re:Mostly Irrevlevant

[0123456]

Don't kid yourself, the much larger issue is the fact that we need autonomous cars in the first place, because distracted humans are doing a pretty fucking shitty job of responding to an "infinite number of scenarios" and killing tens of thousands of humans every year as a result.

We don't 'need' autonomous cars. At current rates of progress, mass-produced autonomous cars will arrive just about the time that telepresense has made cars obsolete.

Self investigation?

[P3r1$c0p3]

"In regards to the self driving car accident, we have investigated our selves and found us not liable. The occupants were. They were taking selfies which is inappropriate and caused the car to crash"

Only because human life is cheap

[pseudorand]

What this really tells us is how the justice system values human life.

Say someone is killed by a self-driving car in a way that's obviously not an hard-to-avoid accident but a clear malfunction of the device. One might expect the liable party to face fairly astronomical damages for designing and marketing a killing machine. But we know they won't. They'll say "That guy made $30k/year and was 10 years from retirement. Here's $300k. We're even." And the courts will say, "ya, that sounds fair.". Maybe a few percent tacked on for the family's pain and suffering or something. But in general, I bet the courts will screw the little guy on this one.

Re:Only because human life is cheap

[Joe_Dragon]

And that thinking will have be like it's better to kill some one then to leave them alive with long term medical care costs.

Self-Insured

[Firethorn]

Lawyers will probably be lining up for contingency fees to go after the corp.

Assuming that Volvo does the 'smart' thing and retains an insurance company to act as a *processor* for claims, they might not be so ready to line up. Volvo would be able to show, in most cases at least, that a reasonable payment offer was extended. This tends to limit punitive damages, which is where they can really make their money.

At least until Volvo has enough self-driving cars to justify having their own claims office and people.

Re:Insurance Companies Are going to Go Batshit

[Firethorn]

Only until it's noticed that said 'drivers' tend to not get in accidents they have to cover as often as even their best drivers. After that, the bidding war to steal the drivers with other companies starts.

There's too many insurance companies for it to not happen.

Re:Will the CTO be willing to due hard time it a c

[Obfuscant]

If CEO of a corporation decides to donate corporate money (not his or her own, but the shareholder's money) to a political candidate, they are effectively speaking on behalf of all the employees and shareholders of the company.

In CU, the "CEO" was using "corporate money", which was provided to him by the shareholders explicitly for that use. If they hadn't wanted it used that way, they wouldn't have used it to form the company. In the case of trade unions, most, if not all states, have laws that say that union members who do not want their money used for political purposes can withhold that part of their dues used for political purposes. This ignores the huge social pressure put upon anyone who does so, and ignores the difficulty in accurately assigning union expenditures to political or non-political purposes. This should be a reason for those who oppose "corporate speech" to rail against unions that do this all the time, but few people who were leaping on the anti-corporate speech bandwagon dared oppose unions who do the same thing.

That makes a huge assumption that all the shareholders and employees agree with the CEO.

That's why there are already laws against such things. Those laws do not remove the rights of the individuals involved in corporations.

Ms car with clippy

[Joe_Dragon]

After looking at your new documents and not making a choice there.
http://vignette4.wikia.nocooki...

I now have some new ones.

Train

Cliff

High speed chase

Crash the gates of local army base.

I'd trust them a touch further

[Firethorn]

Fact of the matter is, providing comprehensive insurance as part of the package for buying a new car is actually a thing over in Europe. They're generally economy shitboxes that a person could probably stop with a hard shove, but a lot of new/bad drivers end up buying said new cars because it's the cheapest option - over even buying a used one because the insurance costs are so high otherwise.

Liability for a new driver overwhelms the expense of everything else. If the self-driving cars have half the *average* accident rate, it'll be an OOM better than the 'new driver' accident rate. So self-insuring will be quite affordable even without playing games.

After that, consider that once a person is used to using a self-driving car that it'll be like automatic transmissions in the USA. They'll tend to stick with self-driving. Successful upsale!

How is this an issue?

[subanark]

Even without self driving cars, it is quite possible that you can be found to have 0 liability for injuring someone:

It was dark, you were driving with your headlight on the highway. As you turn the corner a small kid is out on the street chasing after her ball. You slam the breaks, but you still hit her. Your car was in full working order, and you reacted as fast as reasonably expected. Good chance that the judge finds no one liable, or maybe the parent of the kid for letting them be in a dangerous situation.

You are driving along, and hit the break at a stop light. Your breaks fail and you get into an accident. You've had a recent checkup, and you took all reasonable steps to ensure a safe car. Maybe the manufacture is responsible, maybe the last mechanic you saw, maybe no one.

The only case I see for someone being liable for an accident using self-driving car is:

1. Not keeping your software updated. It would be like not responding to a car recall.

2. Using unauthorized software,, beta software, or software that isn't compatible with your car (including modifying your car).

3. Operating the car beyond the safe operating parameters. Like running the car in extreme weather (in this case the car should detect this and pull over or not start unless the user enables an override, which may be needed in case of emergency).

Re:How is this an issue?

[flippet]

It was dark, you were driving with your headlight on the highway. As you turn the corner a small kid is out on the street chasing after her ball. You slam the breaks, but you still hit her.

In which case you were driving such that your stopping distance exceeded the distance that you could see to be clear, either round the corner or beyond the reach of your headlights, so you were going too fast.

Despite the fact that people get away with flying round blind corners assuming the other side is clear all the time, you are definitely liable on the occasion that it isn't.

This will happen when computers see pigs fly.

[burtosis]

Exactly this.

As a matter of fact, the computer will know about the problem long (hundreds of milliseconds) before you see it and will already be reacting.

The idea that you could react faster or make a better critical decision than the computer is sort of funny actually.

Like those web crawling bots that so easily fill in the captcha boxes? Matter of fact they must be filling it in before i even start to process what is even happening! More like it will be a true revolution when sensor fusion algorithms work half as well as distracted drivers. We are still decades away from even writing algorithms capable of interpreting patterns half as good as an average Joe.

A Red Dward episode?

[easyTree]

Each time someone suffers a broken bone, the damage is mysteriously directed to a member of the Volvo board?

Awesome. Even more impressive than self-driving car tech. You should focus on this as your core product.

Re:Then drop BS claim: self-driving cars humans

[amiga3D]

It's not hard to be better than a significant portion of today's drivers. I look forward to the day I can get in a car and just ride to work while reading a book and not having to worry about the fucking morons that don't know what a turn indicator is or that they are supposed to turn on their lights in a driving rain. The ones that blow by me in the other lane then swerve in front of me to turn into a gas and grocery quick stop. The one's that turn right from the left lane right in front of me. The ones that like to ride 6 feet off my bumper when I'm doing 65 in a 55 zone and I'm still going 15mph too slow for them. The one that ran the light and I just managed to dodge in the intersection and saw them tapping at the phone propped on their steering wheel. Yeah, the bar isn't that fucking high.

I expect self-driving car insurance to be cheaper

[nichogenius]

This may take some time, but after the technology proves itself, the cost of insurance on self-driving models, or even models that employ automatic emergency accident avoidance, should start to go down considerably. Volvo just seems to be shouldering the initial risk factor. Honestly, I don't think I would ever be happy with a fully self-driving car (I like driving my standard transmission vehicles), but they will be a lot better at avoiding those common human error accidents. Unfortunately, getting off the beaten path might be a pain. Imagine trying to backseat drive when a computer is behind the wheel!!