U.S. tells Google computers can qualify as drivers
reuters.com
reuters.com
Out of the 1000s of instances of Joe running around, in the wild, a handful have gotten into situations where he has had points penalized against his license following existing state law. Joe exceeds his points limit and his license is suspended after 2 days.
... This could be interesting.
It's basically an AI collective, whose 'brain' scatters across data centers around the world.
That's a nice picture of free-form learning from inputs, but the reality could be quite more rigid.
I'd expect something more like gathering "black box" style information of the situation (the analogous of a stack trace or core dump), and having programmers trying to improve the algorithms, while also taking into account all the safety issues that could go wrong, than some "let's change my behavior because I was penalized" neural-net/ML way.
Not a lawyer and not a founder, but as I understand it, it goes something like:
- If I own a company and it is insolvent or incapable of fulfilling some contract, I am probably not in trouble, the company is. It might still depend on the specifics, though.
- If I own a company and it does something illegal without my knowledge or intent, such as say produce unsafe products, I might be in trouble, or I might not be, depending on how unsafe, why and whether I can be found have been negligent in allowing it. The company itself definitely is in trouble, though, since it has clearly violated the law.
- If I own a company and I willingly task it, or even allow it, to do something illegal, with full knowledge of the illegal actions in question, then I am probably in trouble too.
Operating through a company doesn't make me immune from the law, and neither should working through an algorithm. On the other hand, full personal liability for everything my company/algorithm does, when I am not in the loop at the point the decision happens, seems like the wrong standard as well.
So basically the algorithm can be found guilty of a crime. The company that created it can then be found guilty of selling a defective product (depending on whether there was any reasonable way of catching the defect). The owners can be then charged with negligence or something to that effect if the defect is something that standard industry practice or compliance with regulations should have caught, or if it was introduced deliberately.
If someone has a law background here, feel free to let me know if I am misunderstanding anything...
That might be technically true but in reality it quite often does. See nearly all bankers from 2K8 melt down for good examples of how money and power buy immunity in what is really an economically based caste system.
Do you really think if this was some garage start-up and not Google that they would have been granted anything but a swift kick in the ass out the office door of whatever bureaucrat handled this?
If the purpose is to account for freak events where the algo misbehaves for reasons that couldn't be foreseen, is that any different than a (non-autonomous) vehicle part failing despite being well-designed and maintained? This is not a new situation - an algorithm is just another car part. You don't need to give legal personality to the fuel injection pumps or the steering column in case it fails unexpectedly.
This issue is only confusing to the extent this isn't considered.
>an algorithm is just another car part
Exactly. And the designer is on the hook for negligence in the usual way - but not for malicious, tragic, and/or unforseeable circumstances that may lead to death or injury.
Finding the boundary between those will keep a lot of lawyers very busy. But I don't think we need to start assigning personhood and agency to a steering system just yet.
A corporation is a legal person, which makes it a lot easier to enter a contract with one, but that doesn't imply it has rights. You can own a corporation without it being slavery.
It's still an abstraction over something akin to mechanical failure, like how corporate personhood is an abstraction over a group of people. It doesn't mean the algorithm has rights, any more than a fuel injection pump does. You also can't murder a corporation, even though it's a legal person.
I'm pretty technically literate, but I have to admit I'm at a complete loss as to why this project matters (let alone what it does). Could you enlighten me?
Let me try.
The Ethereum vision encompasses three components: P2P messaging, a file distribution layer and Blockchains. I described them in more detail here: https://www.reddit.com/r/ethereum/comments/3npsoz/ethereum_i... In practice, this is actually more complex, but should suffice as an introduction.
The Ethereum developers are mostly focused on the Blockchain component. On-chain accounts can be associated with program code and state, the environment specifies a turing complete virtual machine, called Ethereum Virtual Machine (EVM). Transactions can call these programs (also called "smart contracts") and contracts can call other contracts as well. To avoid infinite loops, every execution step in the EVM incurs a fee ("gas cost") that is collected by the miner of the new block that include the transaction. If a transaction does not supply enough gas (paid in Ether, the networks currency), the fees are collected anyway and all changes are undone. The validity of the program state is guaranteed with a consensus algorithm. It is replicated across all nodes. So if you want to censor the program, you have to attack the entire network. This makes them reasonably independent of human control and thus allows them to control arbitrary state, including the networks currency.
But how can the contract communicate with the rest of the world? It can only do so using transactions. For example, it could incentivise transactions that supply it with information, or could use other contracts to schedule micropayments to some service.
Contracts could also be used to reference content addressed data (have a look at ipfs.io, these are basically the other two components) and thereby act as a (domain) name system.
All in all, this decentralises storage, state and communication and is the most futuristic thing I know of. It took me weeks to understand how these components relate to each other, but it was worth it.
Apparently corporations can have religious beliefs in the US.
I have, at least, misgivings about the ultimate decision - and certainly disagree with the religious beliefs claimed by the owners of Hobby Lobby - but I think we should deal with the actual arguments.
(BTW, who's going to pay speed fines? They shouldn't happen but I'm sure they will)
Conversely, what if the corporate shareholders were primarily algorithmic traders?
Human decision makers -- either individuals acting on their own behalf or humans acting as decision-makers in a corporation -- may choose to use algorithms to make decisions, and may choose to give programs that concretely instantiate algorithms minute-to-minute control of their resources. That doesn't mean that the personhood -- of the individual or corporation -- transfers to the algorithm or the concrete instance.
> Conversely, what if the corporate shareholders were primarily algorithmic traders?
Algorithmic traders are legal persons -- either corporations or individuals -- using computer programs to make and execute trading decisions.
This simply seems to be (without the text of the letter, just several news reports) an interpretation of law which is a basis for NHTSA to subsequently rewrite vehicle safety equipment requirements to support fully autonomous vehicles at the federal level.
Because of the way legal definitions are context-specific, it doesn't necessarily mean anything to anynof the other laws regarding drivers, much less anything outside of that domain.
So if anything, wouldn't any sort of sufficiently global optimisation only veer the ai away from behaviours resembling road rage over time?
Most passenger cars are far more capable than average drivers can exploit, so there is quite a bit of room for cars to drive very aggressively from human's perspective, but completely 'safe' from machine's perspective.
"Tests indicate most efficient driving technique to include lots of honking, flashing high beams, and spraying winshield wiper fluid onto cars following too close on the highway."
http://isearch.nhtsa.gov/files/Google%20--%20compiled%20resp...
... along with an abbreviation I have yet to see previously:
>> self-driving vehicles (SDVs)
... and their definition of a Level 4 vehicle:
>> Google describes its vehicles as having what NHTSA's May 2013 Preliminary Statement of Policy Concerning Automated Vehicles calls Level 4 Full Self-Driving Automation. According to that Statement, a Level 4 vehicle:
>> is designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip. Such a design anticipates that the driver will provide destination or navigation input, but is not expected to be available for control at any time during the trip. This includes both occupied and unoccupied vehicles. By design, safe operation rests solely on the automated vehicle system.
"SDV" has been in use since at least 2011: https://scholar.google.com/scholar?cluster=10460039591121716...
The level classifications appear to date from 2013: http://www.nhtsa.gov/About+NHTSA/Press+Releases/U.S.+Departm...
"SDV" is not an acronym but an initialism. TIL that in today's thread about TIL: https://news.ycombinator.com/item?id=11069206
No, it's not really worth mentioning, unless you're implying that Google is blackmailing government employees to expedite rule changes. If you are seriously suggesting that, own it.
Hoover would be jealous of any tech companies, from Apple (remember the Fappening?), Twitter (how many Carlos Dangers are out there?), Mobile Operators to your garden variety ISP, but most of all - the NSA. In fact, if he were alive and in charge of the FBI today he'd wrangle up an ongoing "inter-agency information sharing" set up, and would not be jealous of anyone at all.
Edit: "Our new Privacy Policy makes clear that, if you’re signed in, we may combine information you've provided from one service with information from other services."
I, personally, wouldn't be shocked if a company lobbying for something with potential controversy (Uber, an e-Cig company, whatever) used the data at their disposal to advance their cause.
It might start short of blackmail, sure. My e-cig company is going to face a panel of government officials. Might I check my sales records to see if any of them are customers, to help identify someone on the panel that might be more sympathetic? If I had access to other data, might I push farther?
It's not that I think Google is particularly more evil than any other corporate entity. It's the scale that scares me. Aside from the products I mentioned, there's also Google Fiber, a fleet of camera equipped cars, the Nest home products, and more. And not much in between all that and something scary other than corporate policy and personal ethics.
The biggest, slowest sticklers to this progress thing are at the federal level. This is a huge deal.
I honestly don't get the outrage towards bicyclists. I think it's more psychological? Personally, I have found the most out of shape individuals complain about cyclists the most.
When I see guys on fixies weaving in and out of traffic and blowing through stop signs I worry that all that progress is going out the window. These cyclists are making the roads more dangerous for themselves.
I've almost hit bicyclist blowing stop signs multiple times. While they don't have the mass of a car you can bet that in the case of an injury there's a high probability that the driver is still going to be on the hook for insurance + damages.
If that same biker, when driving a car, runs through red lights then I think that is a more condemnable hypocrisy.
Now, obviously, in an accident, a car will, on average, do more bodily and property damage than a bike. So auto safety is a bigger concern.
this is a tired argument that lacks any basis in fact. what is supported by fact is that we annually kill tens of thousands of people each year in this country (USA) as a result of driver incompetence, i.e. distracted, drunk, etc, etc.......
Yet, I have seen a bike riders pass several parked cars to flat out fly though mid cycle red lights. Drive down the center line passing slow cars on both sides. Do 20+MPH down a sidewalk past driveway exists with limited visibility etc etc.
That's way past 10%.
Bikes are required to take a full traffic lane and obey traffic laws, or use a bike lane. Anything else is both illegal and begging to be hit.
PS: I have personally seen 2 minor bike-car accidents both times the bike was breaking the law. [Car making right hand turn bike passed them on the right side, car pulling out bike hit them on sidewalk]. And a host of near misses where the bike was breaking the law every single time.
Sure, cyclists violate the law, but let's get our priorities straight.
Wait really? This happens to me several times daily in NYC. Hell, yesterday evening it happened going the wrong way on a one-way road.
It will be very noticeable, the autonomous vehicles will be at least an order of magnitude better!
People come up with all sorts of crazy failure modes for autonomous vehicles. What if the sensors fail open at the same time another car is speeding and the sun is at a 37 degree angle and a family of ducks is crossing the road.... Meanwhile the current gold standard can't be counted on to stop when the vehicle in front stops.
Adam Curtis' documentary becomes ever more relevant:
https://en.wikipedia.org/wiki/All_Watched_Over_by_Machines_o...
A quick Google so far reveals:
https://www.americanprogress.org/issues/civil-liberties/repo...
http://www.counterpunch.org/2015/06/19/corporations-over-peo...
https://en.wikipedia.org/wiki/Corporate_personhood
Political contributions and lobbying, things like the TPP, TTIP, the IMF, etc. etc. are all pretty much set up to think about businesses over human lives.
EDIT: This is a good one...
http://www.npr.org/2014/07/28/335288388/when-did-companies-b...
>The massive reduction in risk that is inherent in the development of the modern corporation has been far from fully appreciated.
-John Kenneth Galbraith
As shown in 2008, globalized corporations are able to receive nationalized bailouts even though they are not subject to the same national restrictions of citizens.
Though they don't always get away with it, large corporations such as Lehman are able to commit massive fraud and face no jail time.
Other multinational corporations like HSBC are able to launder money for drug cartels and terrorists but face no jail time.
See also the LIBOR interest rate fraud. No jail time for anyone involved.
Bookkeeping fraud is just the tip of the iceberg. If you are interested in more nefarious activity, like that of the human rights abuses of Shell Oil which runs Nigeria, there is plenty of information out there about it. Slavery, environmental destruction, torture, execution and when any attention is brought to it, the usual punishment is a small fine which is far less than the money made by engaging in these activities.
I guess you just need a little motorized foot that can control a little brake.
And thus high-speed trading was born.
That said, dear Google, please don't take away my foot brake. I'd really rather not have the algorithm decide that the lives of the occupants of my vehicle are worth less than %{UPCOMING_SITUATION} and kill everyone without the occupants having control.
And yes, some people sometimes have to make the difficult decision between hitting another smaller car, hitting a truck, and hitting a person. I suspect these days, most people react irrationally in split seconds. However, with self-driving cars, all the data would be available and a real decision could be made...but what would be the constraints and what would you maximize? that is the scary question.
1) decide that a critical error has occurred and shut down
2) apply full brakes on the remaining wheels, sending the car into a spin and killing a group of pedestrians on the pavement.
3) let the car continue, smashing into a minivan with a family inside it. The minivan is also auto-controlled but has no time to change course
4) try steering further left, hitting the barriers/trees, possibly killing its own occupants.
Now, situations 2-4 can be simulated before they happen with enough computational power. Therefore, the computer would have to take a conscious/programmed decision who to kill, and that is a huge deal morally, because at some point someone somewhere will have to write code that will literally be evaluating value of human life.
And before you say that humans would do no better in this situation - that is true, but no one will judge you for a decision you took in the split second you had, you would do what felt right at the time. The computer is programmed a certain way and you can always sue saying that if it did X then person Y would still be alive - and if there's one thing that companies fear it's litigation.
In your scenario, smashing into another vehicle will be preferable to mowing down pedestrians; that vehicle's program just might be able to save itself, or the two vehicles might even communicate to minimize the damage to each other, and in any case a vehicle will most certainly absorb more of the impact than an exposed pedestrian could.
People on foot, who have nothing to do with the traffic, should never have to deal with the vehicles and the consequences of their programs. Number 2 from your scenario should not even be an option to the program.
Otherwise, can you imagine the first incident in which some passengers died because the code prioritized the life of someone else? Would anyone want to buy an autocar after that? "YOUR CAR COULD CHOOSE TO KILL YOU", the headlines will say.
The second priority should be the safety of pedestrians who are off the road.
Until we have computational power at the level of divine prescience, I don't think you can do much else in scenarios like the one you described, other than have a chain of basic priorities that falls back on "stop everything."
Eventually though, powerful-enough AI may be able to simulate all outcomes down to a very granular level, such as the difference between outright killing someone or just injuring their limbs to the point of crippling them for life.
Many factors will need to be considered then; who appears to be the weakest individual among the unavoidable targets? Are there any acquaintances of the owners among them? Is there a hospital nearby that could tend to their injuries? Does it look like a suicide attempt?
I can imagine that after the first couple of accidents where the algorithm decides to kill N pedestrians to save 1 driver, the government could switch to a "utilitarian" point of view - minimize the number of lost lives - and will enforce this by law.
I made the argument that machine ethics aren't a real thing. Companies will do whatever they can to avoid litigation, then the government will step in with regulations as the closest approximation we'll get to ethical behavior.
There's nothing "utilitarian" about this, unless you assume that:
(a) All people value their own lives equally
(b) Nobody places any value on anyone's life but their own.
(Side note: we know that both of those don't even come close to being true.)
Without those assumptions, a utilitarian point of view will show some pretty obvious biases against just killing the fewest people -- among other things, it will try to kill older people rather than younger people, and to kill people with the smallest number of relatives (who would be sad about a death in their family).
More seriously, I don't understand why this is always brought up in the context of autonomous cars. How often do crashes happen where killing people is inevitable but you get to choose who dies? How often will such crashes happen when autonomous vehicles are common? How far from the theoretical optimum will it be to just say "in an emergency, brake to a stop and steer to avoid obstacles"?
I'm pretty sure this is an edge case on the edge cases. If autonomous vehicles deliver on even 10% of their promise, safety will go way up. It doesn't have to be perfect.
So what do you put in the "Run over the kid, or run into a wall (or oncoming traffic)?" decision point? Somebody has to die. Just braking until you hit the kid is a pretty crappy hack.
And when the kid gets hit, the software will be examined. And the comment that says "screw it, just hit the kid" will come to light.
This is already going to be vastly superior to human drivers, who often don't bother to brake, or flinch and steer into oncoming traffic for no particularly good reason.
After we've cut down the 30,000/year death rate from car accidents by a couple of orders of magnitude, if we get start to get desperate about improving safety and have trouble figuring out how, then maybe we can start looking at rare and bizarre occurrences like these, if they ever actually happen.
Trying to work out the precise requirements for ethics in autonomous car crash response, when we don't have more than the vaguest idea of what kind of crashes they'll get into or what kind of responses might be available, is awfully premature.
And the Google car can already identify a person in the road. That's critical, at least for lawsuit purposes. "You mean, it knew there was a kid in the road, and it did nothing?!"
And lawsuits will happen, the first year. Nobody cares about the vast improvement to humanity; they care that a Google car kit their kid. There's where it all fails.
Computers are poor at interpreting visual information, but can react essentially instantaneously and can pay full attention to all of their inputs indefinitely.
Nearly all crashes are due to those human failings of long reaction times and short attention spans. Even crashes due to equipment failure tend to be greatly exacerbated by those failings.
Autonomous car crashes are going to be due to unrecoverable equipment failures, software bugs, insufficient sensor data, and sudden unpredictable changes in the environment which can't be avoided even with instantaneous reactions.
I don't think it's strange to think that the nature of autonomous crashes will be quite different from the nature of human crashes. I'm certainly not stating that insincerely. I mean, fully a third of traffic fatalities currently happen because human drivers deliberately degrade their own senses and reactions before driving. Another big chunk are because of drivers deliberately not paying attention to driving.
The confluence of factors that need to come together to even create one of these ethically difficult scenarios is so unlikely that I wouldn't be surprised if it never comes up at all. If it does, it's going to be single digits per year. And what's the lawsuit going to look like? Alleging that some Google car should have killed a nun instead of the plaintiff's child? I don't think the law even supports this idea of "you should have taken action to kill someone else instead." Even if it did, it's not going to ruin the industry.
Cars do no just flip over and explode for no reason. Life is not a fast and furious movie.
No offense, but that statistics show that humans are actually terrible at emergency decision making.
Just look at San Diego traffic when it rains. It's a disaster.
Adding even a small amount of algorithmically controlled cars is going to be a huge win.
The bigger issue is going to be traffic and road rage generated by cars actually obeying all the stupid traffic laws.
Cars already ignore driver braking input when it would be dangerous - that's what ABS is, after all, selective releasing of the brakes on locked wheels.
I can think of plenty of circumstances where a passenger just slamming on the brakes could put a self driving car in more danger.
Still, no reason not to have a 'pull cord in case of emergency' type brake for passengers to use to signal the car to stop as soon as can be safely managed.
Most people I talk to don't realize Google has been working on this for a decade (or more) already and the tech is actually really good. Also, safety is a common point many people mention. When I mention how unsafe human drivers are, it's interesting to see how many people never consider that angle.
I'm hoping my current car is the last one that I own. That would require the whole self-driving-Uber thing to come to fruition within next 5 years, but I'm hopeful.
Who state law governing things like driver responsibility would treat as the driver of a fully autonomous vehicle -- and whether they would allow such a thing at all -- is a separate issue.
Level 0: manual.
Level 1: some automation, maybe radar cruse control. (available now)
Level 2: smart automatic cruise control plus lane keeping (Available on several high-end cars now. Tesla is at this level, not Level 3)
Level 3: automatic driving good enough to handle ordinary driving tasks and route planning. (Google, Cadillac/CMU have this in test) but driver sometimes has to take over manually.
Level 4: full auto, all road conditions, driver not needed. (nobody really has this yet)
Now, the NHTSA is discussing modifying the federal safety standards so that, when Level 4 is achieved, vehicles which achieve will comply with federal motor vehicle standards.
Whether to allow autonomous vehicles on roads is a state matter. California DMV currently allows this in test, with manufacturer test drivers only, and requires reporting of incidents. Current DMV thinking is to stay with that for three years, see how things are going, then reevaluate. There's some whining about this from Google, but realistically, Google doesn't have the technology to go beyond that yet.
The accident reports are on line.[1] Almost all of them are from Google, and most of them involve someone rear-ending a Google car which was driving cautiously. Except for one incident.
Cruise (YC W14) had a crash with a parked car last month, driving on 7th St. in San Francisco.[2] The report indicates that the vehicle swerved to the left under automatic control, then corrected to the right, then the driver took over manual control, and then hit a parked car at 20MPH. Both vehicles damaged, no injuries. The reported location [3] is across from the SFPD's Southern Station and a popular parking place for police vehicles. They hit a parked Toyota Prius.
I've written before about the "deadly valley": automatic driving good enough that the driver stops paying close attention, but not good enough to drive reliably. Cruise just demonstrated this. You cannot rely on the human driver to suddenly take over from the automatic driving system.
[1] https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/auton... [2] https://www.dmv.ca.gov/portal/wcm/connect/bc21ef62-6e7c-4049... [3] https://goo.gl/maps/DoNdYSV8cgq
Car driving definitely contains elements that we learn as kids (seeing things in motion, identifying what is a human and what is a rock) while some things are learned as adults (what does that sign mean, how to count KM/H, how transmission works).
Are we really close to autonomous vehicles on the streets?
Except for protests from truck drivers of course.
Close as in "doing proofs of concept and being adopted for experimental tasks?" Yes.
US Gov: +1 Google: +1 Tesla: +1 Uber : -1 Robots: 0xffffffff
Then again, they've contributed to about 200 members of Congress, several senior White House positions are filled by Googlers...
http://www.opensecrets.org/orgs/toprecips.php?id=D000022008&... Congress, 2014 term. Many of the top recipients actually wrote letters on Congress letterhead begging the EU not to rule against Google, which is a fairly unprecedented action.
http://blogs.wsj.com/digits/2014/09/04/two-googlers-take-top... http://www.govexec.com/management/2014/09/white-houses-roste... Googlers take up some of the top technology positions in the White House, including "Chief Technology Officer". Eric Schmidt and other high ranking Googlers also visit the White House nearly weekly.
Think they behaved themselves? Nope: http://techpresident.com/blog-entry/white-house-deputy-cto-a...
In case you're curious how Google's using their White House positions: It's to get money. http://thehill.com/policy/transportation/265932-obama-pledge...
And some more of our taxpayer dollars. While this article carefully avoids mentioning Chromebooks, they're now the leading purchased portable computers in schools, they're pictured in the header of this blog, and it's written by their "former" Googler, who is the CTO: https://www.whitehouse.gov/blog/2016/01/30/computer-science-...
That's a lot of strong government ties, but are they worried about regulation? Well... for the longest time they controlled the US's regulatory body too:
http://www.buzzfeed.com/williamalden/ftc-commissioner-appear... The FTC was basically run by a paid Google shill, Joshua Wright, for the last couple of years. (The FTC reopened their investigation against Google within a month of him stepping down last year.) His pre-FTC behavior wasn't much on the ethical manner either: He was paid as a university professor to write "academic studies" supporting Google's legal views.
One would wonder by the United States has been one of the slowest and most hesitant countries to regulate and pursue legal action against Google, until one does their research. Then it stops being much of a surprise at all.
I mean sure he is biased, but the link doesn't really seem all that negative to be honest. Of course it will have issues in the early stages.
Urmson, of course, whitewashes Google's progress, it's his job. But if you look at the number of miles between safety-related takeovers, or these 'almost accidents', and compare to the estimate of human safety provided by insurance companies, it's hard to believe these cars are yet safe enough to trust on the road. The only reason they haven't been in an accident is because trained, professional humans have intervened.
Self-Driving Cars will probably, eventually, be safer than humans, but Google's PR team has misled people on how far off that point is. Last year Urmson indicated that he wanted to have them ready for consumer use before his kid got his driver's license in four years.
That's pretty unrealistic. Don't believe little old random Internet commenter me? Take it from Robert Scoble, who's been hyping self-driving cars for the last year: https://medium.com/@scobleizer/don-t-worry-uber-lyft-drivers... He's been to see or ride in a variety of self-driving car prototypes from a number of companies working on the problem.
Well yeah, but is anyone expecting them to be put onto the roads in the near near future?
There's already people advocating that California's laws requiring self-driving cars be driven by licensed drivers are potentially discriminatory, and Google's already complaining that their statistically-more-dangerous-than-people drivers currently require licensed drivers on the road: http://www.bloomberg.com/news/articles/2016-01-28/self-drivi...
Given where we're at, it's insane to consider California's laws anything but reasonable. We're a long way off from the necessary confidence to allow a driverless car to run without human intervention. But Google's constant indications that this law and others like it are a problem insinuates a much higher confidence in their technology than any reasonable person should have.
Frequently in these sorts of discussions I've often seen people talking about 2018 or 2019 as likely consumer release dates, which seems to be the real incredible claim needing defense.
> In 10 of the 13 simulated contact events, the SDC’s predicted behavior would have, in simulation, caused contact (though 2 of these involved simulated contact with traffic cones).
The 13 events happened in the 15 months from September 2014 to November 2015 (both incl). 424,331 miles driven.
Can the average driver drive 30,000 miles without making contact with a traffic cone or another car?
The broken pylons at my closest bike lane and highway exit indicate that very few drivers can achieve this.
Detail:
I'm speaking in generalizations here, but it seems that small and even less traveled roads that have pylons often have these damages.
Scarily, on a run today, a "lightly" traveled road in a residential area had multiple "stop for pedestrian" signs that had been knocked down. I find this particularly scary on the (non major side street) that I live - where other than residents (300ppl) and 1 business (< 300 ppl) few people should choose to drive - loses these road pylons in a matter of days.
I am asking, if anything, that we take a more critical look at the statement being made every time Chris Urmson opens his mouth. Because when he brags about the safety of his cars, he's ignoring or failing to mention these incidents where humans saved him from having a very, very bad day.
I agree with some of what you have said, but this is the most bizarre statement I have read today.
There are probably hundreds (or more) traffic cones hit for every small child hit by a car, just because there are way more cars per day passing close to traffic cones than to children.
I personally have had my car hit by others 5 times without reporting it to my insurance company (car was parked, hit and run), as the damage was less than my deductible, and I opted to just live with a few dents. Estimating the number of times a car collides with anything per mile driven is very hard.
My intention in comparing a traffic cone to a small child is the actual safety issue. If a SDC could hit a cone, it could also hit a small child. They're small road hazards which are not generally going to be mapped by the car in advance. I don't want to dismiss the fact that the SDC would've hit a road obstacle, just because in that case the obstacle was a cheap plastic object.
I think the point is that if you counted these kinds of incidents in human driver statistics, your miles per accident numbers would come way, way down.
I suspect those average numbers are being skewed by the rural drivers who rack up lots of miles without interaction with other cars.
San Francisco averages 6.4 years between accidents. https://www.mainstreet.com/slideshow/most-car-accident-prone...
That matches roughly with my experience. And, while the Google car isn't driving San Francisco, the Santa Clara area isn't that much more benign.
But then, we can also go into their statistic basing on a single insurance agency, unreported incidents not factoring in, there's definitely a lot of margin of error here.
I'm glad we're talking about it critically though, and not just glossing over the figures.
One of the other potential points is something Tesla said about Geohot's garage-built car. That it's easy to build a car that performs correctly 99% of the time, but it's exponentially harder to build a car that performs correctly 99.9999% of the time. The number of miles may be less material than even I've focused on here.
Google likes to tout the mileage figure, but it ignores the times that it's human drivers take over. The reality is, driving on straight stretches of road is easy. 99% of driving is easy. Figuring out how to evaluate how safe you are in the other 1% of that time is perhaps an even more challenging problem.
> Google [...] determined that the driver prevented the self-driving car from making contact with another object 13 times.
This phrasing suggests a notably lower threshold of failure than, say, "prevented a crash" or "prevented injury" or "prevented damage".
Thanks to the human driver they managed to avoid the accident. /s
But in all seriousness the damn car was waiting at a traffic light and yet according to you it's still a terrible driver.
There's an additional likelihood that the odd/non-human behavior of Google's cars can make other drivers incorrectly predict their behavior, and also cause more accidents, but I'm not really factoring that into my consideration.
Come to Southern California when it rains and lets see you say that.
The difference is that autonomous safety will continuously improve while human safety has actually been going backwards.
Seriously though, you're also missing the fact that AFAIK, Google doesn't/can't drive these cars in rain. According to a Slate article in 2014, they also can't drive when it's too sunny, because it messes with the sensors. It goes without saying it's never seriously been tested in snow. From my understanding, self-driving cars are mostly tested during the day in fair weather, when humans are also at their safest as drivers. http://www.slate.com/articles/technology/technology/2014/10/...
Also, cars are getting safer, fatalities from cars is actually going down. I managed to find a version of this article that is less blatantly liberal propaganda, because I want to talk about the auto statistic: http://www.lowellsun.com/opinion/ci_29325302/cars-getting-sa...
> Some objects are hard to figure out, though. Heavy rain, for instance, confuses the cars. Google postponed The Bee’s first planned test ride because heavy rain had left puddles in the road, and the cars aren’t sure how to react when the car ahead splashes water into the air. If confused, the cars are programmed to just pull over and stop.
http://www.sacbee.com/news/local/transportation/back-seat-dr...
> “The day I’m very excited about is the day where we are just as safe as human drivers,” he said.
That's a distinct recognition that Google is aware their cars currently are not.
https://www.washingtonpost.com/news/post-nation/wp/2016/02/0...
No, I am fine with alligators walking the planet. I admit, my initial response is a bit stupid. The point I am trying to make is that a.) new technology brings new dangers and b.) there will be people who would like to use these as a weapon (for fun, to terrorize, to make a statement ...)
It would be stupid to outright deny new technology because of this. However you should be informed (learn about all the pros and cons) when deciding on whether or not to adapt a technology.
For instance, so many things have become painful (for good or bad reasons) for the sake of countering terrorism. So why go through all the trouble to finally just offer new toys to the bad boys?
I don't mean everything that can be used as a weapon (which is literally everything) should be forbidden but I don't see a serious debate about the benefits and perils of self driving cars and thus I don't think decision makers can make an informed decision.