On Google's self-driving car acident rates

http://ideas.4brad.com/google-releases-detailed-intervention-rates-and-real-unsolved-problem-robocars

by gwern • 10 years ago

165 181 10 years ago

ideas.4brad.com

Google cache: http://webcache.googleusercontent.com/search?q=cache:9UNd6-l...

Fuzzwah• 10 years ago

I'm going to use this as a place to mention how ridiculous it seems to me that drupal displays the database details on the error page:

the username is drupal and the database server is localhost

I'm sure this isn't a huge information leak, but it still seems kind of silly to me.

Practicality• 10 years ago

While I am not familiar personally with Drupal, this seems like a development feature that is supposed to be turned off in production.

SomeCallMeTim4• 10 years ago

I used to use Drupal. At this point I never, ever recommend people use it.

Pregenerating most of your pages a-la Jeckyll is just so much better that I don't know why anyone ever thought generating all of their pages with PHP was a good idea.

I remember thinking it was a good idea, to be honest. But at this point I look back and see how mistaken I was.

awqrre• 10 years ago

It's always a pain to link to Google also, they leak information in the address that you copy from the address bar...

joe_the_user• 10 years ago

How safe does a self-driving car have to be before it is allowed on the road?

I would say that's going to be a societal decision rather than an engineering decision.

An interesting point is that the introduction of the private automobile itself was an imposition on the social space of that time which basically was wildly unsafe with automobile accidents still being on of the leading causes of death in this country.

If it's determined that self-driving cars will be allowed, ordinary drivers will be forced to adjust to their presence and will have to learn their quirks.

Certainly, cellular phones realistically are being used by a significant fraction of drivers today and if the accident rate has gone up at worst only slightly due to that (not enough to counter-weight other safety measure), it's because non-cell users have adapted to the presence of the cell-user, however annoying that might be.

rhino369• 10 years ago

How safe does a self-driving car have to be before it is allowed on the road?

I'd argue that self driving cars just don't have to be safer than average drivers. They should be safer than average drivers with computer driving assistance. It's already happening, but we can take the self driving programs and use them with human drivers. To get a best of both worlds scenario. Of course we'd have to mandate all new cars come with computer driving assistance.

The counter argument is that we accept the risk of cars now, why not let self driving cars have the same level of risk. My answer is that the risk of cars is so great we either shouldn't be accepting it or are only accepting it because there isn't an easy alternative. A lot of Americans die on the roads.

maxerickson• 10 years ago

It depends on the distribution of drivers that cause incidents. If the average (mean) driver is involved in an average number of incidents, then fine, base the safety estimation on them. If the average driver is involved in a below average number of incidents, the safety standard should be tuned to the drivers that are involved in incidents.

vvanders• 10 years ago

That's a good point, there's been some recent stats(don't have them on-hand) showing AEB making serious improvement in collision and injury rates.

stormbrew• 10 years ago

I think there is a point where assisted driving no longer becomes practical, and either the evolution of autonomous cars has to stop to allow the driver to have a continued say in how the car reacts to its environment or move on to where there is just no place for time-sensitive human input any more. As drivers trust their cars more they will pay attention less (not to mention practice less, an even deeper problem for post-autonomous-car generations than those before), and their ability to meaningfully react to stimulus will be roughly nil.

If we chose the former we are probably arbitrarily limiting our safety in order to save the ego of a human driver. If we chose the latter we will reach a point where we won't have a steering wheel for a driver to potentially become a hazard rather than a help.

cfcef• 10 years ago

Do we expect computer assistance to make much of a difference? Right now they handle the easy tasks like highway driving which don't account for many accidents, as far as I know.

davnicwil• 10 years ago

You make an incredibly good point - whilst any safety improvement at all is obviously and unquestionably a good thing, we can take this opportunity to step back and analyse the situation we've found ourselves in, and decide that the status quo is far from good enough and we should in fact be aiming to do much, much better.

A convenient advantage of regulating for self-driving vehicles is that we can put minimum safety standards into law quite easily, and enforce them meaningfully - because the prize for reaching and adhering to those standards is so great for the players involved, and because we have a relatively clean regulatory slate because of the clear step-change in the technology.

As well as the most important factor, safety, it could also be possible to step back and consider the other disadvantageous side-effects of the current status quo with cars, particularly in cities - noise, traffic congestion, wide roads with thin pavements/sidewalks for pedestrians, indeed a general culture of cars having priority over pedestrians in various situations where the number of pedestrians is much greater than the number of people in the cars.

These are all things which, depending on the city and the culture, can be really significant problems and which I'm sure would never be tolerated had they not crept up on us over decades. Regulating both for and with the capabilities of self-driving vehicles might give us opportunity to vastly improve the status quo in these areas, too, in a relatively short period of time.

It could be quite important to take advantage of the opportunity to realise such optimistic targets now, starting green field, as it can only get more difficult and become much slower to implement after the first set of regulations has been put into effect.

Zigurd• 10 years ago

I'd bet they will just happen one day. "Super cruise control" will get an update that automates your whole trip, et voila.

aresant• 10 years ago

When it comes down to differentiation what do I care about if a robot is driving me around?

Safety. Period.

The equation to get to safety is sensors + computing + map / map geometry.

The sensor differentiation will matter for a while, some OEMs will do very well, but ultimately it will just boil down to the software.

So how is the future of this space not just "everybody licenses Google's platform"?

If life/death safety is the differentiator (and measurable) I feel like the writing is on the wall. Google is going to dominate this.

Maybe Musk sneaks in there with his brash approach of using his drivers as test subjects but his position as a competing auto manufacturer seems less compelling than supporting an "Android-like" effort from Google.

toomuchtodo2• 10 years ago

Maybe Musk sneaks in there with his brash approach of using his drivers as test subjects but his position as a competing auto manufacturer seems less compelling than supporting an "Android-like" effort from Google.

I'd rather support Musk; this is core to his product. For Google, its just another moonshot.

Tesla also has the advantage that Tesla owners drive 40 million miles/month. That's far and beyond what Google's autonomous vehicles are covering (<75K miles/month).

ocdtrekkie5• 10 years ago

A big point in Tesla's favor is that they're putting it out incrementally. Cruise control is a very simple form of 'self-driving' that everyone is comfortable with the existence of.

For the population to truly embrace this trend, it's going to be a long road of incremental steps. Lane assists have started to become commonplace, as have tools to automatically brake to prevent collisions. As these technologies become commonplace, people will accept them. Eventually, fully autonomous driving won't be much of a stretch.

imh• 10 years ago

What kind of info are the Teslas beaming back to HQ? It can't be the ultra high volume data google's collect, can it? This [0] says google's collect almost 1GB/s. Presumably they keep a good chunk of that. Does anyone know how much Tesla is collecting on it's cars?

[0] http://www.kurzweilai.net/googles-self-driving-car-gathers-n...

ocdtrekkie5• 10 years ago

If everyone's going to accept a single standard (which would be helpful for inter-car telemetry), it can't be controlled by Google. We can't let our cars become another one of Google's monopolies.

We should also strongly consider the merits of open source with many vendors contributing and a good governance model. Ideally there'd be some sort of foundation (or government-sponsored group) in charge of such an effort.

The last thing we need is to have our screaming metal death traps buzzing around controlled by black box software by a single developer.

schwap• 10 years ago

So how is the future of this space not just "everybody licenses Google's platform"?

Well I certainly can't see that situation going wrong at all.

At this rate how long until Google re-brands itself as Omni Consumer Products and turns Mountain View into a nation state?

0xffff2• 10 years ago

At this rate how long until Google re-brands itself as Omni Consumer Products

You mean like Alphabet? I only lived in Mountain View for a few months last summer, but I gathered at the time that the city of Mountain View is already fighting tooth and nail to avoid becoming Google City.

dperfect2• 10 years ago

The equation to get to safety is sensors + computing + map / map geometry.

The equation to get to safety is actually what you mentioned + roads where 100% of vehicles are tied in to a common network orchestrating each vehicle's movement in relation to the others and the environment. Without that, differentiation in software is likely to lead to unanticipated dangerous scenarios, just as the HFT market is susceptible to flash crashes... except on the road, they will be real crashes with the potential to be more destructive than human-caused accidents.

Hopefully when that network becomes a reality, we can build it on top of some kind of open, distributed platform (perhaps related to blockchain technology) that no single entity can abuse.

eric_the_read• 10 years ago

How does a single network 100% of vehicles are tied into avoid the problem of an attacker only having to subvert one node on that network to compromise everybody? (Not to mention the difficulty/impossibility of getting to 100% in a reasonable timeframe.) Seems like that's been shown time and again to be practically impossible.

jessaustin3• 10 years ago

That seems really brittle. The baseline for every vehicle should be safety in every situation. (Sometimes that might mean: pull over to the side of the road and stop.) Only once every vehicle is proved safe, can certain combinations of vehicles negotiate compromises to safety in pursuit of other goals like speed.

jameshart2• 10 years ago

"Android-like?" You mean allowing manufacturers to ship crapware-infested versions; having manufacturers handle deployment of critical updates, meaning they simply don't; having core features overridable via aftermarket appstore downloads that provide a vector for malware?

dragonwriter8• 10 years ago

having manufacturers handle deployment of critical updates, meaning they simply don't;

The consequences for a manufacturer of failing to deploy a critical update for self-driving cars is substantially different than failing to deploy a critical update to a mobile handset. So even with a similar set of formal, contractual requirements, there are reasons to expect that the resulting behavior could be quite different.

wsetchell2• 10 years ago

I can see this market playing out to be an oligopoly not a monopoly.

- Very high barrier to entry. You need $X of development costs, and $Y on traffic/mapping data. Anyone with money can get those. - Self driving cars have limited economies of scale.

  * Self driving taxis get better (less wait time) as their density goes up, to a point. Today Uber and Lyft are both dense enough in SF for me to not care which one I use.

  * Self driving cars get safer as you collect more driving data. Safer self driving cars are better up to a point. Safety matters a lot now, since cars are unsafe. On the other hand I don't care if Boeings are 2x as safe as Airbuses, both are safe enough for me. If you can make your self driving car safe enough, I'll buy it. To make your self driving car safe enough, you need to spend $X on developement and testing and $Y on really good maps.

magicalist2• 10 years ago

So how is the future of this space not just "everybody licenses Google's platform"?

I feel like what we've already seen in the DARPA grand challenge and in the self driving car research world will be what happens in the future: trailblazing is really hard work, but once others can see it can be done, even without a lot of details of how it was done, doing the same thing is usually much easier, or at least more obviously viable and so worth the investment.

Not to mention all the published research, suppliers with established manufacturing, and experienced engineers looking for other jobs that would now be out there.

Starting a self driving car company ten years from now, say, still would not be remotely easy, but you won't need anywhere near twenty years to catch up.

willvarfar• 10 years ago

I can't wait for self driving cars. End of that discussion each time we eat out about who is going to drive us home...

DanFeldman• 10 years ago

The CA dmv also publicly reports all accidents involving autonomous vehicles [1]. Most accidents are caused by drivers taking over control manually, or by other actors on the road driving erratically (or rear-ending the self-driving vehicles). The entries on the page report Google, Delphi, and Cruise automation as having incidents, though they're all minor damages, if there are any at all.

[1](https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/auton...)

huangc105• 10 years ago

I live in Mountain View and I see Google self-driving cars around all the time (day and night). My personal feel is that they look "different" and my attention will deliberately focus on them (which may or may not affect my driving when I'm around them).

Why not test with cars that look "normal" (think early 2000s Corolla). Wouldn't this further decrease the chance of possible accidents?

I guess what I'm saying is, who's decision was it to make it look like a toy car and not just an actual regular car that doesn't divert my attention off the road?

*edit for better readability

hueving• 10 years ago

Marketing. You want people to notice and think, "Hey, that's one of those self-driving cars, that's cool!"

huangc105• 10 years ago

I believe you're correct in that marketing did have a strong sway in the decision.

But the fundamental principle behind Google has always been engineering and especially for something as crucial as safety, they should have stuck with something bland and nonchalant imo.

grillvogel3• 10 years ago

google is only capable of designing things that look they were made by and for 5 year olds

huangc105• 10 years ago

True statement for some Google products but not necessarily all.

a_c_s• 10 years ago

In this case the more non-threatening the cars are, the more likely people are to accept them.

themoonbus• 10 years ago

True, my 5 year old cousin loves using google inbox to improve his productivity.

jimrandomh• 10 years ago

They want to get people used to the idea of sharing the road with self-driving cars. People will be a lot more comfortable with them and have a lot fewer misconceptions if they've seen and recognized enough of them for it to stop seeming noteworthy.

huangc105• 10 years ago

Is this really the case though? Wouldn't you want to produce results, facts, and data first?

My sense is, prove it's safe first, then mass produce something that IS safe in the design you want.

After all, once something is mass produced then people adopting it would be quick. No?

epaga• 10 years ago

*accident

underbluewaters3• 10 years ago

I think it's pretty clear that the bar for safety must be much higher (10x) than human-level for these to be accepted by consumers. Auto accidents are very common, and the first time someone has an accident with a car like this everyone they know will hear about it. It will be terrifying. If these are only as safe on average as a human driver then nearly everyone is going to have at the very least a 2nd-hand negative experience.

The emotional response to these accidents is not going to be entirely irrational either. If I have a minor accident with my traditional truck, I'm going to probably have a good understanding of what went wrong and how I can prevent future collisions. With an autonomous vehicle... software upgrade? I'd rather take responsibility for my own safety if that's the case.

thrownaway24243• 10 years ago

The idea that humans learn well from accidents stands contrary to all available evidence.

ajkjk2• 10 years ago

What evidence? Certainly individual humans (anecdotally: like myself) do.

ignoramous• 10 years ago

And what evidences are you bringing to the party?

underbluewaters3• 10 years ago

Of course I'd learn from each fender-bender, and I'm clearly a better driver than average anyways! These fancy autonomous vehicles are too dangerous for my family.

ethbro• 10 years ago

"There were 272 instances in which the software detected an anomaly somewhere in the system that could have had possible safety implications; in these cases it immediately handed control of the vehicle to our test driver. We’ve recently been driving ~5300 autonomous miles between these events, which is a nearly 7-fold improvement since the start of the reporting period, when we logged only ~785 autonomous miles between them. We’re pleased."

785 miles between anomalous events that require manual control is bad enough. 5300 is worse! That honestly needs to be 0 events (impossible as that might be) or only events for which advanced notice can be provided to be market-acceptable.

It's the TSA screening problem all over again: 13-88 hours of mind-numbing normal operation, with a single incident of something potentially exploding that you have to catch.

drdeca• 10 years ago

You mean that 5300 is worse because people will pay less attention?

Well, would it be better if it had a random chance of (artificially) needing human intervention, at on average every 785 miles? (Not exactly, because that would make it too easy to predict, and therefore might not make the person attentive when a real problem happens)

That doesn't seem like it would be too hard to add, if it was better by keeping people on their toes.

But, I doubt the problems get monotonically worse when the average distance between events increases? I would expect at some point, increasing distance doesn't significantly reduce attention , so the safety would increase a bit. Maybe not more than for some shorter distance, but I think for some slightly different distance.

Xixi3• 10 years ago

Which one will learn faster: each individual driver each time they have a minor accident (or a close-call)? Or the autonomous vehicle software updated from data collected from each and every accidents involving an autonomous vehicle? The result will probably be software-updates like Tesla cars are getting today.

I suspect that if the attitude towards autonomous vehicle accidents is similar to that of aircraft accidents, they will become extremely safe very quickly.

Silhouette• 10 years ago

I suspect that if the attitude towards autonomous vehicle accidents is similar to that of aircraft accidents, they will become extremely safe very quickly.

On the other hand, if every autonomous vehicle accident is investigated to the same standard as every aircraft accident, it appears likely that the entire system will become so overloaded and cost-ineffective that it is completely impractical to operate before we get over the hurdle and achieve any consistently safer performance that might otherwise be available.

fullshark• 10 years ago

The point is not which is safer on a bottom line level, the point is consumers are terrified of the idea of surrendering control to a computer. Lots of people are terrified of flying for example even though it's much safer. Unless it's 10x safer people will be scared of automated cars and possibly even push for short-sighted laws.

wrsh07• 10 years ago

I agree that the bar for safety should be ~10x human-level. But I also find it interesting, because that should be a moving target. Eg if we realize humans often rear end the car in front of them, and fix that with "auto braking", then we'll improve how safe human drivers are.

Thus, hopefully, both humans and self-driving cars should have improving records [although I don't have any data to back this up].

J-dawg2• 10 years ago

You've raised an important point here: the 'human vs self-driving car' is something of a false dichotomy.

The human driving experience will become more and more augmented by technology, with auto-braking, lane-assist, smart cruise control and (eventually) auto-navigation until we have almost-self-driving cars which are still legally required to have a qualified human in the driver's seat. Many of the these features already exist in high-end cars.

People will be much happier to hit the 'auto' button and climb into the back seat when they've already had several years of 'almost-auto' driving.

magicalist2• 10 years ago

I think it's pretty clear that the bar for safety must be much higher (10x) than human-level for these to be accepted by consumers.

Will it? Think about the recent Toyota brake pedals that didn't work or the GM faulty ignitions. Both of these caused a number of deaths, and the Toyota one was especially scary because it didn't even need accident-like events to cause a problem, your brakes would just stop working.

It'll hit the news (like Tesla battery fires vastly disproportionately reported on over other, much more common auto fires), but it's not clear at all that it will cause some massive panic or backlash.

doall2• 10 years ago

About the Toyota's unintended acceleration (UA) incidents. From Wikipedia: "On February 8, 2011, the NHTSA, in collaboration with NASA, released its findings into the investigation on the Toyota drive-by-wire throttle system. After a 10-month search, NASA and NHTSA scientists found no electronic defect in Toyota vehicles.[27] Driver error or pedal misapplication was found responsible for most of the incidents.[28] The report ended stating, "Our conclusion is Toyota's problems were mechanical, not electrical." This included sticking accelerator pedals, and pedals caught under floor mats.[29]"

https://en.wikipedia.org/wiki/2009%E2%80%9311_Toyota_vehicle...

If this was true, then the self-driving car may have prevented the accidents. If it was really a software problem, then fix the software. It may be much more easy to fix the software rather than to fix human behaviors.

caskance• 10 years ago

I find it unlikely that consumers are as safety conscious as you imagine them to be. Think about how many people currently text or do other things to distract themselves while driving. They willingly choose less safety in order to be able to do something else while in transit. Pushing this behavior to its natural conclusion is what will allow self driving cars to be accepted even if they are 10% less safe than normal drivers.

underbluewaters3• 10 years ago

People text while driving because they think they are better at doing two things at once than they actually are. For the same reason, they will expect that they can out-perform a robot unless those robots are so overwhelmingly safe that conclusion can't be rationalized. No consumer will be thinking about the big picture.

rain206• 10 years ago

I think in this case, insurance companies will help drive adoption. If an insurance company believes that accidents are say half as likely, you would have to imagine one of them would bite on changing premiums by a bit. Cold hard cash will facilitate adoption, assuming basic regulatory guardrails are in place.

zhanwei• 10 years ago

I think it could be predictability of a self-driving car in an accident that is making people uncomfortable and accident rate does not capture that. For e.g., an accident might means hitting a barrier, a human realizes the terrible situation and will swerve away from the cliff behind the barrier. But what would a self-driving car do? Its sensor might be too damaged to realize the dire consequence.

bsder• 10 years ago

I think it's pretty clear that the bar for safety must be much higher (10x) than human-level for these to be accepted by consumers.

That's a really low bar.

Take a look at Google Maps in San Diego when it rains. Accidents everywhere. A self-driving car won't get into stupid accidents because it's following too closely, driving too fast, hydroplaning, etc.

Toss in the people who are distracted, tired, or drugged out on medication, and your numbers are even worse.

We're looking at the end of humans driving cars within 10 years.

brownbat• 10 years ago

I think it's pretty clear that the bar for safety must be much higher (10x) than human-level for these to be accepted by consumers.

I think the safety level just has to exceed that of an average teenage driver for widespread adoption by a certain segment of the car-buying population: parents.

EDIT: Teen drivers are 3x more likely per mile to be involved in an accident than adults - http://www.cdc.gov/MotorVehicleSafety/Teen_Drivers/teendrive...

draw_down• 10 years ago

I think it's pretty clear that the bar for safety must be much higher (10x) than human-level for these to be accepted by consumers.

You might be right about that, I just don't see why it should be the case. Humans are not very good at driving safely, it seems like an easy bar for robots to clear. Even if robots are "only" twice as safe, we should all jump at the opportunity to halve injuries and deaths from auto accidents.

bduerst• 10 years ago

How does that number compare with humans? Well, regular people in the USA have about 6 million accidents per year reported to the police, which means about once every 500,000 miles.

Aren't accidents only reported to police if there is a hit & run, or some other criminal activity?

Seems like the wrong metric for comparison, given the way they define the self-driving "accidents" and that the majority of human fender benders are not reported.

cheald• 10 years ago

Not necessarily. Major accidents will have the police involved, small fender benders might be handled with just an exchange of insurance information. If there's substantial damage though, the police will probably get involved since a police report makes collecting on insurance easier.

HCIdivision17• 10 years ago

It's not uncommon to contact the police so that a quick statement can be made and the paperwork logged. It seems to help with the insurance stuff and getting a third-party's eye on the scene.

dsp1234• 10 years ago

In some(most/all?) states, all accidents over a certain dollar amount or that caused injury to any party are required to be reported to the police.

But the lack of police reports is why the article mentions that insurance companies believe the accident rate is higher than the 'official' number.

Anecdotally, the number of times that someone has hit my car, and then asked me not to report it in exchange for cash is non-zero.

iamleppert• 10 years ago

All of the brouhaha about self driving cars is moot until real user testing begins. I'm not talking about driving a google mobile around the streets of mountain view with an engineer in the driver's seat.

These vehicles (I would hope) are designed to drive around people in different circumstances. They're software for use by people and until we have realistic tests with different kinds of people, ages, driving experience levels, etc. it's all academic.

I'm somewhat wondering if all the self driving car stuff by Google and Tesla is primarily a vehicle (no pun intended) for marketing rather than actual tech that will yield a real product.

bertil3• 10 years ago

It seems, from previous reports by Google, that their ability to understand discrepancy between sensors relies on knowing the road very well, not just the map with nuances like “cars can go there, but it’s mainly a pedestrian street” but more local undocumented habits. They mentioned details that prove that they would have to drive a lot in new cities to learn enough and make their car safe.

The one that I remember is that they have started driving in Austin and came across a very local species of vehicle with a unique habit: the hipster fixie rider with his stand-stop motion at red lights. Committed bikers do not have free wheels and stay standing in place not by putting their feet down but rocking back and forth; that motion was new and not interpreted clearly by the car, that hesitated to see it as a false start. It’s since been fixed. The article didn’t specify if the shirt pattern had any meaningful weight in the interpretation.

abalone6• 10 years ago

What about manufacturer liability? That's the critical question here.

It can't just be 2X or even 10X safer than humans if the manufacturer is liable for the accident. That would EXPLODE the cost of these cars and bankrupt companies -- even if just a small fraction of today's fatal accidents due to driver error became Google's liability.

Consider that even single-vehicle accidents, which today might be the driver's fault, could result in a manufacturer lawsuit.

This is why I think fully autonomous without any human to intervene is a pipe dream, because the safety level required for that is pretty close to perfectly bugfree in an almost infinitely complex world of roads and conditions. Just for liability reasons.

These test cars of course have human copilots. But it remains to be seen if that really will grant manufacturers immunity for accidents. If the car is perfect 99.999% of the time, wouldn't that train the human to trust it and not pay as close attention? And then miss that 0.001% of the time when it hurts someone? Would a judge find it reasonable for a human to stay vigilant and responsible for that fatal corner case bug that cropped up after two years of perfect autonomous driving?

Liability is The. Critical. Question.

mikeash2• 10 years ago

Are you saying this because manufacturers, being gigantic unsympathetic companies, would be sued for much greater damages than at-fault drivers get hit with today? Otherwise it makes no sense, TCO will come out the same whether drivers or manufacturers are liable.

fweespeech2• 10 years ago

It can't just be 2X or even 10X safer than humans if the manufacturer is liable for the accident. That would EXPLODE the cost of these cars and bankrupt companies -- even if just a small fraction of today's fatal accidents due to driver error became Google's liability.

Right now, the insurance company is liable for the accident followed by the consumer. It'll just shift the insurance market towards manufacturers buying it [or likely self-insuring] and the cost of insuring a car will be baked into the price instead of bought aftermarket.

dragonwriter8• 10 years ago

Right now, the insurance company is liable for the accident followed by the consumer.

Actually, the driver is liable, but in most states (in order to be legally permitted to drive on public roads) must either have purchased insurance to cover some minimum amount of that liability or have posted a liability bond. But the insurer's obligation to pay is triggered by the driver's liability, not ahead of the driver's liability.

DenisM• 10 years ago

There is no explosion. The cost of collisions today is reflected in the car insurance, so about $100/mo. If the manufacturer were to absorb this liability at current incident rates, it would make cars about 30% more expensive (assuming $300/mo loan payments). If Google reduces the (incidence * impact) of collision by 10x, the cost will come down to $10/mo, or about 3% of the total car cost.

And in return for the modest price increase you get a self-driving car!

abalone6• 10 years ago

You're forgetting about tort damages. Insurance does not typically cover the cost of killing or seriously injuring someone. Those damages go into the millions, far exceeding the maximum coverage. Cases of serious injury and death often end up bankrupting the responsible individual.

Also consider that there are more people who can sue per accident, because humans aren't responsible anymore. Today if you're driving the car and get in a accident and it's your fault, you can't sue anyone. Now you can. That's one more party per accident, and a whole category of single-vehicle accidents than now result in lawsuits.

snowwrestler8• 10 years ago

The cost of collisions today is reflected in the car insurance, so about $100/mo.

Not true--consumer auto insurance premiums do not encompass the cost of consumer lawsuits against manufacturers, because of course manufacturers do not buy consumer auto insurance.

For example Toyota settled their unintended acceleration lawsuits for $1.2 billion dollars or so. GM's ignition switch problems cost them $900 million to settle. How much will Volkswagon's diesel fraud cost them?

None of these costs are covered by consumer auto insurance. Instead they're covered by the capital of the corporation, which ultimately comes from investors or customers.

Now imagine a self-driving world in which every accident opens up the possibility of such class action settlements. In the absence of legislation to cap damages, companies would have to set aside huge pools of money to defend and settle such lawsuits. That means less investment, lower returns, or higher prices--any of which tend to hurt the long-term viability of a company.

maratd• 10 years ago

More than likely, insurance will continue to be mandatory and paid by the vehicle owner ... but seeing it drop to $10/mo would be pretty sweet! Of course, you would then be prohibited from driving the vehicle. Self-driving only if you want to see the savings.

aidenn0• 10 years ago

There is no explosion. The cost of collisions today is reflected in the car insurance, so about $100/mo. If the manufacturer were to absorb this liability at current incident rates, it would make cars about 30% more expensive (assuming $300/mo loan payments).

Except juries invariably award much larger damages against companies then against individuals; particularly when the individual is dead or permanently disabled by the accident. People will be reluctant to levy a multi-million dollar award against a reckless driver's widow, but will be perfectly happy to award it against a multi-billion dollar corporation.

dragonwriter8• 10 years ago

It can't just be 2X or even 10X safer than humans if the manufacturer is liable for the accident. That would EXPLODE the cost of these cars and bankrupt companies -- even if just a small fraction of today's fatal accidents due to driver error became Google's liability.

If the owner/operator bears no liability and all liability is on the manufacturer, it transfers the liability cost onto the manufacturer, but even if it is as safe (not 2× or 10× as safe), that just means the manufacturer rolls the cost to (self-, likely) insure for that liability into the purchase price (but the purchaser doesn't need to get their own insurance, so the total cost of operation to the purchaser is unaffected.)

And, of course, if a company like Google is both the manufacturer and the owner/operator (e.g., using the vehicles in an Uber-like service tied in with Google Maps and Google Now, or using them for Google Express delivery vehicles with smaller robots onboard to deliver packages to the door, or using them for Google Street View camera vehicles, etc.), operator vs. manufacturer liability makes no difference.

bertil3• 10 years ago

You are assuming that, if liability in the US was a critical question, Google would not launch its project in a country with most hospitable weather and more compatible liability rules and from there, lobby for more reasonable principles in the US.

ergothus• 10 years ago

Or we could learn to stop suing people over everything. I mean, gross neglect is one thing, but if it's a safer driver than I am, and it screws up, we call that an accident.

...

Nah, I'm dreaming, it's far more likely that insurance companies will set up a wide insurance program ala taxi services than it is that we'll learn to stop being litigious.

lucb1e• 10 years ago

gross neglect is one thing, but if it's a safer driver than I am, and it screws up, we call that an accident.

Exactly my thoughts. You typically have very little to go on if a company did something to you and they were not in neglect. At least in Europe/The Netherlands, I don't really know about the US. (The stories we hear is that in the US, neglect means not warning that coffee is indeed hot, or that you should not put pets in a microwave to warm them up... I don't believe that really so I'll just assume it's similar to here.)

petra• 10 years ago

If it's only 10X better than humans ,and considering that professional drivers are about 10x better than the average, maybe we're better of with something like shared UBER ?

Or maybe just add self driving for the highway , especially if it's possible for the driver to get off the vehicle before entering the highway , and hop on another car entering the city ?

danieltillett3• 10 years ago

In theory this just shifts the insurance coverage from the driver to the manufacturer. The real unknown is if any insurance company will cover the manufacturer or not. I suspect that like the airline industry there will need to be a legislated cap on damages to make this work. Provided that there is insurance then there should not be any issue.

More fundamentally we should not assume that the people working on this at Google are idiots. They already know all of this and I am sure they are working with insurance companies and governments to solve this problem. Compared to the technical difficulty of making a reliable and safe autonomous vehicle this is an easy problem.

SomeCallMeTim4• 10 years ago

1. Air bags sometimes kill people who would otherwise have survived in an accident, but on balance they save lives.

2. Making air bags mandatory hasn't caused manufacturers of cars or of air bags to go bankrupt.

3. If self-driving cars cause deaths, but on balance save lives, the situation is directly analogous.

4. There will be a way to make self-driving cars work. Q.E.D.

I seem to remember that laws exist to grant immunity to manufacturers of airbags, at least if they were functioning properly. A quick Google finds lots of ambulance-chasers who talk about "defective airbags," so manufacturers aren't immune to liability if they screwed up.

Similarly, you could imagine that No Fault laws could be passed for self-driving cars that were operating within reasonable parameters. A gross error on the part of a self-driving car could still leave manufacturers liable, but as history has shown, car companies are willing to write off even gross negligence claims rather than voluntarily add safety devices to cars, calculating cost-to-fix vs cost-to-pay-off-family tradeoffs. [1]

In fact, I'd go as far as to say that, any company that doesn't offer self-driving cars in the next 15 years may as well plan to shut down or be acquired. This is going to be a must-have feature. They will figure out how to make it work for manufacturers.

The liability question will be solved. They wouldn't even let car companies go bankrupt when they (arguably) deserved to; they certainly won't let them get into a situation that kills them, and as others have rightfully pointed out, the numbers aren't even that prohibitive if they needed to self-insure.

[1] http://content.time.com/time/specials/2007/article/0,28804,1...

abalone6• 10 years ago

That is a very poor analogy. Airbags have only killed a few hundred people and in many of those cases the manufacturer was still not liable because of human error (victim was not wearing a seatbelt, was a child improperly seated in the front, etc.)

The manufacturer liability exposure for fully autonomous driving is literally several orders of magnitude greater than for airbags.

MagnumOpus• 10 years ago

quick Google finds lots of ambulance-chasers who talk about "defective airbags,"

That is the Takata airbag case - they manufactured defective airbag with malfunctioning ammonium nitrate inflators which exploded supersonically instead of inflating, killing 8 people. They were fined $200m just recently and have to recall and replace 400,000 airbags.

This case might actually be analogous to self-driving car issues - it was a case of a defect potentially caused through negligence in engineering, rather than airbags working as intended.

avar3• 10 years ago

Liability of all things is "The. Critical. Question." when it comes to autonomous vehicles?

That's such an amazingly American thing to say. Do you really think that the uptake of autonomous vehicles is going to be hindered to any significant degree by what are largely U.S.-specific legal issues?

Can you imagine a world in 2060 where the U.S. decides to leave something like 10-20% of GDP on the table because of the liability aspects of their legal environment? It might hinder uptake a bit but I don't see it stopping it in the long run.

In any case regardless of what the U.S. does at that time the rest of the world is going to be driving autonomous vehicles.

abalone6• 10 years ago

In what country are you not liable for hurting someone with a car? I want to know.

openasocket2• 10 years ago

Product liability is hardly unique to the US: you can sue a manufacturer for a defective product in a great many countries.

wsetchell2• 10 years ago

I don't think liability is that key here.

cost of damages = # of accidents * cost per accident

- Self driving cars should reduce the number of accidents

- Cost per accident seems independent of the kind of car you drive. If a human driven taxi hits me, I can sue the taxi company. If a self driven taxi hits me, I can sue Google. I should be able to sue for the same amount in either case.

Animats• 10 years ago

Volvo's CEO: "We are the suppliers of this technology and we are liable for everything the car is doing in autonomous mode. If you are not ready to make such a statement, you shouldn't try to develop an autonomous system."

So there.

At first, self-driving cars will probably be leased on operating leases, with maintenance and insurance bundled into the payments. Or maybe you own the vehicle, but there's a maintenance and insurance contract required to enable self-driving.

bradtemp• 10 years ago

I (the author of the OP) have an article about accidents at http://robocars.com/accident.html looking into many of these issues.

absherwin• 10 years ago

Minor crashes are even more frequent than the article estimates. Thus, the real human accident rate is even higher. Probably between 1 in every 24000 and 87000 miles.

The VTI driving study[1] equipped 100 cars with sensors and was therefore able to measure all crashes experienced. It directly measured 1 crash per 24000 miles. If we extrapolate based on the 17.4% police report rate, that suggests 1 per 87000 miles.

[1]http://www.nhtsa.gov/DOT/NHTSA/NRD/Multimedia/PDFs/Crash%20A...

aetherson2• 10 years ago

That's an interesting study. I'd be cautious at trying to create a one-line conclusion from it. It is, however, a fascinating full read, and not very long and not very technical, so I'd encourage people to read the whole thing.

Note the narrow demographic data and geographic data.

chad_strategic• 10 years ago

Next time you pull up to a stop light, look to the left and then right. I'm willing to be the drivers on the left and or right are looking at their phone.

It's getting worse... the phone, blue tooth, the dashboard computer, traffic, impatient drivers. etc...

I for one can't wait these self driving cars, they have got to better than distracted drivers we have now.

ams61103• 10 years ago

I don't mind someone looking at their phone while they're waiting for a red light. I'm more worried about the ones who do it while in motion.

Thirdegree• 10 years ago

Next time you pull up to a stop light, look to the left and then right. I'm willing to be the drivers on the left and or right are looking at their phone.

And the dude between them is busy staring at his neighboring drivers rather than the road!

fweespeech2• 10 years ago

http://webcache.googleusercontent.com/search?q=cache:9UNd6-l...

Cache. Its down for me.

VLM• 10 years ago

Why bother?

In a world of massive un- and under- employment, mechanical turk and massive connectivity, increasing income inequality, and generationally declining standards of living, a handful of millionaires will sit in their unimaginably expensive self driving car, and five mechanical turk remote drivers will cooperate on a 3 of 5 vote basis to drive the millionaire.

Most brainpower is underutilized, the quantity is increasing, and its manufactured by unskilled labor. There eventually comes a peak-automation point where its better and cheaper for the economy and culture to just hire a dude to do it.

The world of the future is only ivy league grads will have real jobs and they'll have all the money, whats so inherently awful about someone in the favelas getting a job as a driver, especially if its cheaper and safer?

oofabz• 10 years ago

The world unemployment rate is 8% according to the CIA World Factbook. What's going to make this increase so dramatically? We've had industrial automation, computers, and the internet for decades now, and people still have jobs.

Also why would you want someone remote-controlling your car instead of sitting in the drivers seat? Wireless connectivity is way too flaky. Tunnels block the signal, rural areas have poor coverage, and equipment breaks and requires maintenance. It would be insanity to bet your life on a cellular signal.

bertil3• 10 years ago

I am surprised that Google has not started having their software recording on cars that are driven by humans, to learn. They could increasingly have feedback to say “Our software would not drive that fast/that close to that curb” and progressively simulate a lot more dangerous situation and learn about tough situations than they currently can.

occamrazor• 10 years ago

They have a vast number of human operated vehicles with extensive telemetry data: Street view cars.

argonaut• 10 years ago

This is a harder problem than you think it is. This is a classic reinforcement learning problem, and it is incredibly difficult in the real world. It's easy in discrete state-space, perfect information, turn-based games like Go (the recent advance there). Very difficult in robotics.

TazeTSchnitzel• 10 years ago

If the humans always take the wheel in dangerous situations, how do you know the cars are safe? The cases where they are most likely to get into trouble are the cases where the AI is not in operation!

brianmcconnell• 10 years ago

Licensed pilot here (learned to fly in 1987). People need to look at self-driving cars in the way pilots use an auto-pilot.

Auto-pilot is most useful in two situations. 1) long cross country legs where there is not much flying to do (just maintain heading and altitude), so A/P frees the pilot up to manage other systems, enjoy the view, etc, and alleviates fatigue, 2) flying a precision instrument approach, reducing the risk that the pilot will succumb to spatial disorientation in the setup for (usually manual) landing.

With cars, auto-drive capability will be useful in reducing accidents in two modes: 1) long duration highway driving where fatigue is a big issue, 2) intervening to prevent a distracted driver from causing an accident (rear end collision etc).

I'd be perfectly happy with a car that can drive itself in cruise mode on the interstate, but requires an alert driver on local roads (with the added bonus that if I am about to slam into something, it will brake to avoid or lessen the impact).

Something for the liability crowd to consider, self driving cars won't be able to avoid every potential mishap, but they will be able to reduce their severity. A car that can automatically brake to reduce its speed by 25% just before impact, will reduce its kinetic energy by roughly half, and the potential for injury by more still.

crpatino• 10 years ago

This is a very insightful comment, from a person who has first hand user experience with self-navigation systems. Sadly, if you dare utter blafphemy against the Almighty God of Technology, bad karma will happen to you.

This is even against the hacker spirit. We loathe IDEs and cherish emacs/vi close to our hearths because what we want from automation is augmentation of human capabilities, not self mutilation + prostaethics. How can people not accept that the same formula can work best for cars?

pathsjs• 10 years ago

The fact is that auto-pilot cars, as you describe them, won't be nearly as revolutionary as self-diriving cars. The latter would allow people to put to good use the time to commute to work; they would improve car-sharing to the point that people would not need a personal car and parking could become a problem of the past; they would allow you to get back home safely after some drinks; eventually people would not need to learn to drive at all. The mere augmentation of human drivers, as useful as it would be, would not change much in our habits

J-dawg2• 10 years ago

I made a similar comment elsewhere in the thread before reading yours.

I think the evolution of self-driving cars will be through the gradual improvement of auto-pilot cars until the human just has to sit there and do nothing 99% of the time (but still be legally in control).

stillsut• 10 years ago

Your comment made me think: Could driverless cars kill the domestic airline industry?

If I could get from NY to Florida in less than a day in my driverless car, and at sub-$2-gas, who would want to do the security check-in, luggage ripoff, flight delay thing when going less than 1000 miles?

knughit• 10 years ago

Are you familiar with Amtrak.com?

vonnik• 10 years ago

typo in the headline. we need one more 'c' in accident.

tjl• 10 years ago

Is Google only testing in California? It seems that the data for accidents was given for the whole of the US, but it should be compared to California only. I don't know if it's noticeably different or not, but I'd think that snow and ice would increase the accident rate in some states.

Snow and ice I think would provide a challenge for self-driving cars.

illegalsmile• 10 years ago

Turns out snow and ice provides challenges for cars driven by humans!

tim333• 10 years ago

we have to figure out just how to test these vehicles so we can know when a safety goal has been met. We also have to figure out what the safety goal is.

Or you can just crack ahead and try them with users like Tesla.

f137• 10 years ago

250,000 miles on average between accidents? Does anybody think this is unbelievably high number?

dsfyu404ed• 10 years ago

The average human driver is consciously forgoing the option of driving in the hyper-conservative manner that the google cars do...

On Google's self-driving car acident rates

Related Stories