I'm torn on this issue. On one hand you have people who view it as their moral duty to pass on confidential / classified information that clearly shows breaches of the law to the proper authorities. On the other hand, you have the current scatter-gun approach where huge swathes of data is released, some of it unrelated, with the likelihood that at some point in the future something unrelated to the original subject of the disclosure is going to get into enemy hands with very real and devastating consequences.
However, both of these things are trumped by something Mr Blair just doesn't seem to get - if you ask people to keep secrets they find unconscionable, they won't stay secret for very much longer. I'm also greatly concerned that Mr Blair considers insider whistleblowers to be a "new" threat.
The answer is in your second paragraph. Why is the government doing unconscionable things that they are asking people to keep secret? As to the first paragraph, doesn't the government always tell us that if we aren't doing anything wrong we have nothing to worry about? Whenever we have these leaks, they aren't about how the government is doing food drops to help starving people somewhere, no, it is about the government doing things that are in violation of its own laws and treaties.
Lord Blair mentions the damage that Snowden did, exactly what damage? Are we really that ignorant to believe that other governments and terrorist groups didn't realize that the intelligence community was using electronic surveilance? Only the public wasn't aware of what was going on. People can argue all day long whether Snowden was right or wrong in what he did, that's not the point. History has shown that secrecy leads to tyranny.
He warned there was a 'new threat which is not of somebody personally intending to aid terrorism, but of conduct which is likely to or capable of facilitating terrorism.' "
I don't know who we should be more concerned about actual terroriss or guys like Lord Blair? It seems that he has the potential to do much more harm to society. Next thing you know he'll probably want tougher laws on who can have sex because that too is conduct which is likely or capable of facilitating terrorism by producing future terrorists.
Are you going to tell us one more time that rushing designs is not a good thing? You preach the obvious as if it were insightful and/or relevant. It's neither, so unless you are going to create a compelling argument that these cars are being rushed to market too soon you might consider dropping this redundant argument. Hint: presuming that nobody will solve the problem of how to input an address into a computer within the next 20 years is probably not going to help your position.
by Google's estimate, it will take 20 years before there are enough of such vehicles on the road to effectively mitigate the risk from drivers not driving these cars. Don't like the 20 year number, go talk to Google.
Do you have a reference for that? I'm guessing that you are totally misunderstanding or misrepresenting what they've said, as you have in just about every post you've made here.
And if, as you believe, that 1/2 the posters to slashdot could solve the remaining problems Google is facing before going live, well, maybe Google should hang out here more often and pick their brains. One would think, though, that their engineers are better equipped to deal with the issues than the average slashdotter. But who knows, maybe you are right. If so, it's time to unload some Google shares.
Yes, the Google engineers working on the project are much, much smarter than the average slashdotter, but once again you have completely misunderstood what's been said to you. This happens so often it seems like it must be willful ignorance on your part. But to be even more abundantly clear, those Google engineers are indeed going to be hard at work solving the problems that you haven't even considered. NOT the ridiculous problems you've pulled out of your butt that hardly deserve mentioning other than indicating that (despite the mysterious inside info you have which requires multiple NDAs) you have zero technical knowledge on self-driving cars work, taxi services, or car sharing services work.
So I'm not shorting Google, I'm shorting your company. Although I will have to consider the possibility that your business model is just to pretend to be incredibly dumb on slashdot so that other slashdotters will give you all their clever ideas for free.
Reread the original article. While not associated with Uber, you will find insight as to the reason for the NDAs and whether or not I know what I am talking about or not. For record, I know more about self-driving cars and taxi services than you could possibly imagine. Here's a hint. If this endeavour is successful, our investors will make a lot of money. If it's not successful they will still make a lot of money, just for different reasons. You see, when you play your cards right and invest in key pieces of underlying technology required to make things happen, you win regardless.
Too much said already. Go wait for consumer grade self driving cars or whatever. That's not where the money is and Google knows it, at least not yet.
Why wouldn't a car have 360 vision? You also wouldn't need 10 feet. People back into things because they don't see them. Someone diving for the wheel is going to be pretty rare. A kid riding a bike along the sidewalk and the driver who is 6 feet into the driveway not seeing the kid and hitting them with the back of the car that is on the sidewalk is going to be much much more common.
Google's own research shows that a kid riding a bike down the sidewalk is not a problem (so do most accident reports). On the otherhand, their own research shows that backing up while people are standing next to the vehicle is problematic, especially if a young child is present. Should somebody be standing next to a vehicle with a young child while the vehicle is backing up, no. Do people stand next ot a vehicle with a young child while the vehicle is backing up, yes. The car cannot protect against every stupid act. For instance, there are not sensors under the car looking to make sure I am not standing there with my foot behind the tire and if I am, my foot will most likely get run over. Or there isn't an AI program running that is trying to decide if the kid throwing a baseball in the next yard is going to try and intentionally hit the car with the ball. Chances are if he does, depending on where he is throwing from, the car is going to get hit.
But there are sensors that look for kids and animals and obstacles that are behind the vehicle or may be moving towards being behind the vehicle while it is backing up. AI can't solve for every random case (well it can, but it would be cost prohibitive). Instead it tries to solve the most likely scenarios, particularly the ones with the greatest likelihood of injury or death.
That's not a criticism of google or the technology. It's just a recognition that as in everything, you have to make choices.
All of that is non-sequitur to the fact that their rates are not based on good math. The accountants don't care how you determine your price. They just care that the amounts coming in and going out are what has been declared as the amounts that should be coming in and going out.
Almost. The independent auditor definitely doesn't care what the amounts are but does care that they are calculated and recorded appropriately. Insurance companies (and any company that is self-insured) has to post a liability for the unfunded part. So, if the actuarial reports show that you have an IBNR (incurred but not reported) liability and claims in process and projected claims that exceed the reserves created through your funding, you still have to post the liability and associated loss. That loss and associated liability lowers shareholders equity and if it is large enough could make the insurance company (or self-insured company) balance sheet insolvent where they have more liabilities than they have assets. If that occurs, the state could step in because the insurance company can cover its obligations or if a self-insured company, creditors can force the company into bankruptcy.
As such, the accounts and auditors, being independent, don't care what premiums the company charges, however, the CFO and BOD do. Sure, policies need to be priced to be competitive in the market, but not so much that they lose money or shareholder equity per contract written. Corporate officers can be held liable for not exercising due diligence including pricing policies for contracts.
So, it is safe to say that if contracts are being priced without rhyme or reason just because of what some other insurer is pricing them then either a) there is enough excess profit in the contracts that it is not an issue, or b) the risk calculations from the other insurer for the locale being covered are similar enough for the company in question. Anything else would deplete company assets.
Why should the Pope being insulted have anything to do with whether the earth moves around the sun? Why are you making ad hominem attacks against Galileo, and throwing out your own "evidence-free" assertions that he made "evidence-free" assertions? What does someone thinking someone else is an asshole have anything to do with their actual science?
Those aren't ad hominem attacks, the whole Galilelo thing is extremely well documented. I'll grant you the asshole comments, that part is just the original poster's opinion, but even that is pretty well accepted about Galilelo. But, Galileo did refer to Pope Paul V by a derogetory term in one of his remarks, that's recorded in the transcript.
The irony is that the whole thing was not about science. The Catholic Church already accepted the theory of heliocentricsm. What they held was that Galileo's proofs did not actually proof it based on the aristotlelian system that was standard for the day. Galileo went to the church to settle the dispute between him and the scientific community that held the same position. He thought that since the Jesuits agreed with the heliocentric model that the pope would too. But that question wasn't the questioned posed. Instead what the tribunal was asked to decide, by Galileo himself, was whether or not his proof met the requirements to be put forth as a truth instead of a theory. The Church said no.
So, I don't know if Galileo was an asshole, but based on the transcripts he was pretty arrogant. And when he didn't get the answer he wanted the second time, he basically told the pope to go screw himself. The rest is history.
He was also correct about heliocentricism but wrong as to why as he based a large portion on Copernicus' theory for the non observance of parallax shifts of starts which modern science has shown to be wrong (actually as soon as better equipment was available, Copernicus was shown to be wrong in this area). So what is it called when you get the right answer but for the wrong reasons?
Galileo was friends with Urban the VIII, but it was Paul the V that the controversy began (Urban suceeded Paul). Actually, the controversy was not about the actual science but more politics. There were already theories of heliocentricity from both Copernicus and Kepler who preceded Galileo. The standard for science back then was based on an Aristotlian system. In proving his work, Galileo relied on Copernicus, and while heliocentricity was more or less accepted in the scientific community and many in the Catholic Church, there was much dispute about the great distances between starts that Copernicus theorized to deal with the abscense of parallax shifts. The problem for Copernicus was one of crude instruments, but because his theories were not universally accepted by the astronomical community of the time, Galileo, basing his proof on Copernicus failed the Aristotlian rigor needed to for proof.
Galileo disagreed and took it to the Church assuming that since the Jesuits agreed with him, the Pope would, too. But the Pope sided with general astronomers of the day and said that he was free to teach his theory but it was not a proven fact. Luther basically said the same thing to Kepler 10 years earlier, but the Lutherans don't get in trouble for it.
Even after Galileo was placed under house arrest for publishing his work as fact instead of theory (which is what the dispute was about), the Church provided housing for him, built him an observatory, fed him, provided servents for him, paid him to do research and a host of other things. It was probably the most comfortable house arrest in history.
Anyway, there were large egos involved and Galileo refused to change his position and said that he was correct and the Church was wrong. While history has shown his theory to be correct, it is for the wrong reason. Copernicus was wrong on the parallax shifts and if they had better instrumentation he would have seen the shifts. So in a way both sides were correct at the time. The heliocentric model was correct, although that was never really disputed, but the Church was correct in that it failed the rigors of scientific proof.
Only in your head is 20 years off a "rush to production." But yes, many of the problems will take many years to solve, however, not the FUD you're throwing out there. Those "problems" are in some cases actually benefits, others of your "problems" have already been solved and the rest can be solved by half the posters to slashdot if they are not afraid of trying (which obviously you are). What's next, are you going to complain that the animatronic driver is not lifelike enough?
You know, since you brought up the whole man on the moon thing. One only need to look at NASA's AS-204 (Apollo 1) tragedy to see what rushing designs can do. Nobody is saying to sit on these cars for 20 years, but by Google's estimate, it will take 20 years before there are enough of such vehicles on the road to effectively mitigate the risk from drivers not driving these cars. Don't like the 20 year number, go talk to Google. They are actually pretty forthright with the capabilities and limitations of the vehicles. They are trying to protect themselves from the over-hype that is flying around about these vehicles.
And if, as you believe, that 1/2 the posters to slashdot could solve the remaining problems Google is facing before going live, well, maybe Google should hang out here more often and pick their brains. One would think, though, that their engineers are better equipped to deal with the issues than the average slashdotter. But who knows, maybe you are right. If so, it's time to unload some Google shares.
Jeez, Joe. How many dumb excuses are you going to manufacture in a feeble attempt to discredit self-driving cars? Just about every circumstance you mention happens today with regular taxis and current drivers. Do you think NYC taxi drivers today get 100% fidelity in understanding the destinations of their passengers? It's far more likely that the driver himself will have an accent, but even then you are not accounting for the millions of passengers who speak little or no English. The reality is that even with real taxi drivers the destination will increasingly be communicated non-verbally. You punch in where you want to go on your smart phone and the car picks you up, drops you off and you get charged automatically. No struggling to understand which Hyatt you mean, or who gets which cab.
For whatever reason, you don't like change, but please try a little harder to come up with some plausible problems that don't exist today. Just about every post you've made on this thread is filled with faulty assumptions and bad math.
I'm not sure why you think my name is Joe, probably from my nickname, but that is just from a persona from something unrelated. However, I actually embrace change. Particularly, I embrace change in technology fields. However, as any other VC will tell you, it's not all about the technology, but the perception of the technology by the public.
For instance, it is well accepted that betamax was better technology than vhs, but for a number of reasons the best technology did not win. Likewise, when technology is overhyped in what it can or will do, that may do great for initial sales, but it hurts investors, particularly VCs. That's why, if you look closely, Google and Google's engineers aren't hyping up the cars.
Yes, they are a great piece of technology and they have the potential to change transportation. But they don't claim they will be accident free. They don't claim that they will solve all of these various problems that you see people claiming and they even admit they will create new problems. Now why would they do all that? Because they know for the cars to be successful, they need to avoid the hype. The last transportation vehicle that was overhyped was promoted as unsinkable and it sank. Google does not promote the cars as uncrashable. Their own research shows that statistically their cars will be involved in accidents. Chaos theory guarantees it. No, what Google promises is that their vehicles will be less likely to be involved in an accident than a human driven vehicle. There data also supports that but that is a far cry from what people are claiming. They do not claim it will be the safest vehicle. They do not claim it won't prevent every kind of accident. The do not claim it will bring world peace and feed the hungry and everything else. They simply claim that it will be less likely to be involved in an accident than a human driven vehicle. That in itself is still pretty impressive.
All of the rest of the discussion, and you can believe me or not, but they have been going on with Google and VC and potential commercial buyers are involved with the public perception. That is where things like language barriers or using smart phones or body fluids come from. They are all mentioned in Google's investor literature and are in areas that they are working on addressing.
Google is releasing these first vehicles in the commercial market instead of the consumer market. Why? Because it is easier to convince a CEO or CFO of the value in practical terms instead of all of the hype. The technology pretty much speaks for itself. On the otherhand, what commercial accounts will be interested in is the impact to the bottom line and the ROI.
Things like having a car pick you up after keying something in on your phone isn't going to happen anytime soon. Most states don't even allow google's car and those that do require a driver behind the wheel. Eventually those laws will change, but states will be slow to change until the technology has proven
Better watch it, as I have found out on this thread, posting anything even remotely critical of Google's vehicle is a sure way to get modded down. Only time will tell how safe Google's vehicles are. I have a good idea, but several NDAs keep me from elaborating. I will point this out. The goal of Google's cars were never to be accident free. That was determined early on to be cost prohibitive. No, Google only promises that it is safer than human drivers, which is a much lower standard to beat.
This is smart of Google. The last time a vehicle was touted by a manufacture as the safest, the word used was unsinkable and that was a titanic mistake. Google doesn't want to make that mistake. They know it is impossible to account for every contingency and it is likely that the vehicles, once in wide spread use, will be involved in accidents (although most likely they will not be the cause of them). Google's engineers know that even the best AI can't see around blind curves or break the laws of physics to avoid something unexpected. At best it will do better than a human at worst it will do no worse.
If you want accident free driving, it's not going to happen. Chaos theory would indicate that it is impossible. However, if you want self driving cars with an acceptable level of risk (whatever that level may be), then that is possible. That is what Google is developing.
These problems are only solvable if you bother to spend a minute to think about them. Who has time for that? I'll believe self-driving taxis can work the day they're finally able to put a man on the moon.
Of course one can rush to production and skip all of the analysis you seem so glib about and spend your time in court with a bunch of lawsuits or you can spend the time up front analysing what could go wrong and come up with ways to mitigate the risk. The choice is totally yours.
Hmmm, they must have changed the design, but unless it has 360 vision, the most common backing accident with a child is with the child along side the vehicle and then somehow getting under the wheel, whether by reaching for something or falling or reaching for the moving vehicle or any number of scenarios. I guess an intelligent car can just sit and do nothing while a child is within 10 feet of it or turn over control to the manual driver.
No, they are not. I used to write insurance rating software. You know those companies that would offer to compare rates against different companies to get you the lowest price? In California, we were the major supplier of the software that does that. With that, we had very close contact with all of the insurance companies, as their agents would use our software to quote rates.
Contrary to 'common knowledge', auto insurance policies are not well thought out mathematical calculations. They are kludge on top of gut feeling on top of assumption on top of more layers of kludge, gut feeling and assumption. What a large number of the companies do is just clone a competitor's plan, and than change calculations based on what they assume will give them good results. I can't count the number of times I called a company to get clarification on their calculations and the response I got was "I don't know. How are you calculating it? We will just calculate it the same way you do.".
That's interesting because when auditing insurance companies, regardless of what premiums they charge they need to come up with the IBNR and various other claims data and they have to be able to show that their premiums charged and reserves on hand are enough to satisfy the various statutory and accounting requirements. I have no doubt that there is pricing pressure, but, from the actuary and accounting side, the marketing pressure is ignored, unless they are willing to take a qualified audit opinion, which most insurance companies are not. But maybe California rules are different than other states, but that still wouldn't explain how they get around GAAP and the auditing standards.
Your amount really only includes worst case scenario. Plenty of accidents happen where the damage is less than $25,000. I would say the vast majority of people have cars that are worth less than $10,000. Also, $300,000 for liability might be right in some cases, but most of the time that won't happen. I guess in the US it's more likely, as they don't have universal health care, but I still don't see the average payout being anywhere over $10,000. Many accidents will simply go unreported because people don't want their premiums to go up, or get demerit points on their license.
Your insurance company is insuring against the risk associated with whatever you policy states, so if you have a vehicle worth 25,000 and 300,000 in liability, that is the risk they calculate your premium on. If that premium is $900/yr, then the risk of them having to pay out is very low, which is the point I was trying to make.
Your bad assumption is that the vast majority of accidents require paying out the full liability and vehicle coverage. They don't.
Traffic statistics, btw, are 3.1 billion vehicle miles driven per year and around 10 million accidents. Or in other words one accident every 300k miles.
However, that assumes that people only have one accident and yet a few posts up is a person who has had 6 in 300K miles. You can't calculate accident rates that way. My bad assumption is based on the actuarial reports of our own self insured auto fleet. Yes, I over simplified it, but, basically, of the 100 drivers we have statistically less than 10% of them will be in an accident in any given year. Our drivers average about 80,000/yr. So in one year, they drive the equivalent of what the average person does in four. Statistically we would expect 10 drivers to be involved with accidents of some sort this year. Accidents even include another vehicle backing into our parked vehicle (something likely to be the case with a google car, too, and actually the number one reported accident). In addition, the 10% rate is for being involved in any accident. The study shows the same 100 drivers being involved in an accident that they were at fault in is less than 2%.
A 10%/yr accident rate does not mean that in 10 years all 100 drivers will be in an accident any more than a 50/50 chance of having a boy means that every family with two children will have a boy and a girl. Each probability stands on its own for each new year (or birth in the boy/girl example). Some routes are more prone to accidents then others as are some drivers. We have drivers with over 10yrs experience and 1M miles without a single accident.
My point is that you can not evaluate risk or mitigate it with something as simple as taking the number of accidents and dividing by the number of people. That will only yield meaningless numbers. The insurance industry has a whole "science" built around actuarial studies to understand the risk involved with insuring drivers and how to mitigate that risk. While that process is much to complex to deal with adequately in/., it should be enough to state that a handful of google cars not having any accidents is no more relevant risk wise than a handful of Ford Focus' not having any accidents.
Google cars probably are as safe as people are implying they are, but it would be better if there was industry standard data to support that claim. Lack of such data means there is just anecdotal data and that really isn't data at all.
You think a company like Google, that sends cars down every road taking pictures every 10n feet or so where n is small, is relying on your maps, do you?
I don't.
Google says it uses government sources for their maps and traffic data, so, while probably not the paper maps, it probably is the same government GIS data that the paper maps are generated from.
Not that this will stop anyone the first time the car backs over a kid, despite their excellent safety record.
The Google cars have backup cameras, radar, and bump sensors. They have been specifically designed and tested to not run over kids/pets while backing, under many different light and weather conditions. So your scenario is very unlikely to happen.
A much more likely scenario: After self driving cars are common, some human driver backs over a kid, and people ask why we should continue to allow humans to drive.
By the time a bump sensor responds it is too late. As for backup cameras, radar, etc. It is all a matter of physics and how fast the car can decelerate versus how fast the object is moving that just came into its field of so called awareness. Chances are, backing out of a driveway will be slow enough that unless the kid is on something moving quickly, it will stop in time. On the other hand, most often, children aren't hit directly from behind but glance off the rear fender and fall beneath the wheel. It is hard to see how a google car will prevent that.
You simply cannot calculate the chance of having an accident that way. Driving 300,000 miles in Montana is going to be different than driving 300,000 miles in New York City. If you want accident statistic rates, one need look no further that auto insurance. If it was a matter of dividing total mileage by number of accidents, none of us could afford our premiums.
If your articles don't make sense. The first one, the pro google article (what would you expect them to tell you about the safety) claims that people have 1 accident per 165,000 miles driven (not taking gender into question). The second article, even if men drove the same as women, would be 1 accident every 175,000 miles. Averaging the 2 rates given yields a rate of 1 accident approximately every 186K miles and in reality the number would be higher as there are more male drivers than female drivers. But even at the 186K figure that is 13% greater than the first article's number. Now, it has been a long time since I studied statistics, but a 13% variance in something that is measuring a finite set seems that there is bad data somewhere.
In reality, it goes back to Mark Twain's there lies, damn lies and statistics. Without valid samples, statistics are pretty much meaningless. Without knowing the types of accidents, any projections on deaths are meaningless (how many of those accidents are in parking lots or while parallel parking?). If your goal is to reduce fatal accidents, you can do that now, by lowering speed limits. That would be more effective now versus 20 years from now, the earliest projection when google cars will reach enough percentage of vehicles on the road to impact highway safety).
Until then, all we know is the project accident rate for a google car. Your insurance company has that same figure for you, too and they figure your premium on it. If you premium is manageable, then you are not an insurance risk and you are no more likely to be in an accident than a google car. If it is high, then you are. Plain and simple. Insurance companies are in business to make a profit and they are pretty good at analyzing the risk involved with insuring drivers.
I am pretty sure the average person has an at fault accident more often than 300k miles driven.
I'm pretty sure you are wrong or the average person's car insurance would be significantly higher than it is today. Assuming you are correct, more than one would equate to at least 2, so that would be at least 1 at fault accident every 150,000 miles. Full coverage with 300,000 liability is around $900/year assuming you are over thirty in this part of the country. Driving 15,000 miles/yr would mean 1 accident per 10 years, so the insurance company will collect 10,000 from you in premiums and potentially pay out $325,000 in claims (300,000 liability and 25,000 vehicle). If the average person is so likely to have an at fault accident, how would the insurance company stay in business in that business model (which was intentionally over simplified for this example)?
No, the fact of the matter is that most people never, ever are involved in an accident at all. That is why insurance, while not cheap, is relatively so compared to the cost being insured. Think of it like life insurance. A 20 yr old can get some pretty cheap life insurance, why? because the likelihood of the insurance company paying is very small. The same can not be said for an 80 yr old.
Agreed, 300,000 miles without an accident isn't that awesome.
Yes it is! Many of those miles are not cruising the freeway, but on a test track under conditions that were designed to cause an accident. Test dummies have been used to simulate pedestrians stepping into traffic. Other cars pull in front, or cut off the Google car, or drop objects onto the road. The self driving cars have been able to avoid thousands of accidents where a human would likely not have been able to react in time.
So you are saying that the 300,000 miles is in a laboratory controlled environment instead of a real world environment? Boy that instills confidence, because we all know how realistic that is. I wonder how many Boeing 787 had battery fires in their controlled test environments instead of real world environment? I would hazard a guess of not too many or the problem would have been addressed before releasing the plane. Controlled tests can only go so far. Obviously if you fail the controlled ones you will fail the real world ones, but passing controlled tests doesn't mean it will pass real world tests.
Hopefully, you are just posting without knowing and the google researchers actually did most of their testing in the real world and not on some test track. Otherwise, the vehicles shouldn't be allowed until proven safe.
Noun 1. An unfortunate incident that happens unexpectedly and unintentionally, typically resulting in damage or injury. 2. A crash involving road or other vehicles, typically one that causes serious damage or injury.
Accidents can also be caused by chance, but the word itself doesn't have to mean that. When someone says there was an accident somewhere, they aren't (necessarily) implying that nobody was to blame.
I think replacing human driven cars with these things would save a lot of lives.
In risk management there is a big difference between incident and accident. When two airplanes fly too close (what is call a near-miss), that is an incident. If they actually hit, that is an accident. All accidents are incidents, but not all incidents are accidents. What is needed to evaluate the google car is the incident rate, not the accident rate. Why? To minimize accidents, you need to minimize incidents. If google cars are involved in a high rate of incidents, even if they avoid accidents, then the risk of an accident is high.
Think of it this way. Most teenagers do not have accidents, but they do have incidents. Accidents always occur from incidents, so insurance premiums are higher on teenage drivers. It is not the accident rate that is important in evaluating the self driving cars, it is the incident rate. Because even with low accidents, if there are high incidents eventually there will be high accidents.
The google car already has over 300k miles on it without a single at-fault incident. Although I thought the law required a person to be in the car ready to assume control at all times?
I drive an original 1973 VW Beetle every day with over 300K miles on it without a single accident (at-fault or otherwise). I would not use that statistic to say that all VW Beetles are safe vehicles. Just because a small handful of these cars have been tested does not mean that they are safe in the real world. To do a proper study, you have to have a large enough sample size. Maybe that has been accomplished, because that probably only needs around 1,000 vehicles. However, testing those 1,000 vehicles in different locales and conditions would seem to require more than 300K miles to have a statistically valid sample size.
I'm not knocking what google has accomplished, but really, has the car been tested in enough scenarios that it is safe or is it safe enough that the cost of making it safe outweighs the cost of the likelihood of injury (kind of like Ford and the Pinto)?
They will be far less likely to back over a kid, or confuse pedals like oldsters around here love to do. This is because the outside of the car can be covered in sensors instead of being a hinderance to visibility to the driver.
Really? statistically, what is the likelihood of a taxi backing over a kid or even being driven by a senior citizen. These are self driving cars for consumers, these are commercial vehicles such as taxis and delivery vehicles.
As for covering a vehicle in sensors instead of being a hindrance to visibility to the driver, the same visibility requirements exist for human driven vehicles and self-driven vehicles because humans have to be able to drive self-driven vehicles on occasion, so if you need to cover all the windows with sensors, it isn't going to work.
If your main concern is backing over things, they've had dash board cams and backup sensors for years now.
But statistically, it'll probably be better than having humans behind the wheel. Not that this will stop anyone the first time the car backs over a kid, despite their excellent safety record.
Think about that. These are going to be used as taxis in New York. So, not only will you have to be able to get one, but it will have to figure out where you are going, whether you speak with an accent or not. Of course, you could just type your destination into your smart phone and forgo the talking, but if the car can't get the voice recognition algorithm correct, then that makes the safe driving algorithm suspect.
Then who is going stop all sort of stuff from happening in the back seat? Will the next person really want to sit in other's bodily fluid? Or what if two people want the same cab? Will the car choose or will they fight it out and winner take all?
It seems that self-driving cars have fewer complications for individual drivers instead of as a fleet of taxis, but then, you could always still keep a taxi driver in them defeating the point of having them in the first place.
Slashdot Mirror
I'm torn on this issue. On one hand you have people who view it as their moral duty to pass on confidential / classified information that clearly shows breaches of the law to the proper authorities. On the other hand, you have the current scatter-gun approach where huge swathes of data is released, some of it unrelated, with the likelihood that at some point in the future something unrelated to the original subject of the disclosure is going to get into enemy hands with very real and devastating consequences.
However, both of these things are trumped by something Mr Blair just doesn't seem to get - if you ask people to keep secrets they find unconscionable, they won't stay secret for very much longer. I'm also greatly concerned that Mr Blair considers insider whistleblowers to be a "new" threat.
The answer is in your second paragraph. Why is the government doing unconscionable things that they are asking people to keep secret? As to the first paragraph, doesn't the government always tell us that if we aren't doing anything wrong we have nothing to worry about? Whenever we have these leaks, they aren't about how the government is doing food drops to help starving people somewhere, no, it is about the government doing things that are in violation of its own laws and treaties.
Lord Blair mentions the damage that Snowden did, exactly what damage? Are we really that ignorant to believe that other governments and terrorist groups didn't realize that the intelligence community was using electronic surveilance? Only the public wasn't aware of what was going on. People can argue all day long whether Snowden was right or wrong in what he did, that's not the point. History has shown that secrecy leads to tyranny.
He warned there was a 'new threat which is not of somebody personally intending to aid terrorism, but of conduct which is likely to or capable of facilitating terrorism.' "
I don't know who we should be more concerned about actual terroriss or guys like Lord Blair? It seems that he has the potential to do much more harm to society. Next thing you know he'll probably want tougher laws on who can have sex because that too is conduct which is likely or capable of facilitating terrorism by producing future terrorists.
Are you going to tell us one more time that rushing designs is not a good thing? You preach the obvious as if it were insightful and/or relevant. It's neither, so unless you are going to create a compelling argument that these cars are being rushed to market too soon you might consider dropping this redundant argument. Hint: presuming that nobody will solve the problem of how to input an address into a computer within the next 20 years is probably not going to help your position.
by Google's estimate, it will take 20 years before there are enough of such vehicles on the road to effectively mitigate the risk from drivers not driving these cars. Don't like the 20 year number, go talk to Google.
Do you have a reference for that? I'm guessing that you are totally misunderstanding or misrepresenting what they've said, as you have in just about every post you've made here.
And if, as you believe, that 1/2 the posters to slashdot could solve the remaining problems Google is facing before going live, well, maybe Google should hang out here more often and pick their brains. One would think, though, that their engineers are better equipped to deal with the issues than the average slashdotter. But who knows, maybe you are right. If so, it's time to unload some Google shares.
Yes, the Google engineers working on the project are much, much smarter than the average slashdotter, but once again you have completely misunderstood what's been said to you. This happens so often it seems like it must be willful ignorance on your part. But to be even more abundantly clear, those Google engineers are indeed going to be hard at work solving the problems that you haven't even considered. NOT the ridiculous problems you've pulled out of your butt that hardly deserve mentioning other than indicating that (despite the mysterious inside info you have which requires multiple NDAs) you have zero technical knowledge on self-driving cars work, taxi services, or car sharing services work.
So I'm not shorting Google, I'm shorting your company. Although I will have to consider the possibility that your business model is just to pretend to be incredibly dumb on slashdot so that other slashdotters will give you all their clever ideas for free.
Reread the original article. While not associated with Uber, you will find insight as to the reason for the NDAs and whether or not I know what I am talking about or not. For record, I know more about self-driving cars and taxi services than you could possibly imagine. Here's a hint. If this endeavour is successful, our investors will make a lot of money. If it's not successful they will still make a lot of money, just for different reasons. You see, when you play your cards right and invest in key pieces of underlying technology required to make things happen, you win regardless.
Too much said already. Go wait for consumer grade self driving cars or whatever. That's not where the money is and Google knows it, at least not yet.
Why wouldn't a car have 360 vision? You also wouldn't need 10 feet. People back into things because they don't see them. Someone diving for the wheel is going to be pretty rare. A kid riding a bike along the sidewalk and the driver who is 6 feet into the driveway not seeing the kid and hitting them with the back of the car that is on the sidewalk is going to be much much more common.
Google's own research shows that a kid riding a bike down the sidewalk is not a problem (so do most accident reports). On the otherhand, their own research shows that backing up while people are standing next to the vehicle is problematic, especially if a young child is present. Should somebody be standing next to a vehicle with a young child while the vehicle is backing up, no. Do people stand next ot a vehicle with a young child while the vehicle is backing up, yes. The car cannot protect against every stupid act. For instance, there are not sensors under the car looking to make sure I am not standing there with my foot behind the tire and if I am, my foot will most likely get run over. Or there isn't an AI program running that is trying to decide if the kid throwing a baseball in the next yard is going to try and intentionally hit the car with the ball. Chances are if he does, depending on where he is throwing from, the car is going to get hit.
But there are sensors that look for kids and animals and obstacles that are behind the vehicle or may be moving towards being behind the vehicle while it is backing up. AI can't solve for every random case (well it can, but it would be cost prohibitive). Instead it tries to solve the most likely scenarios, particularly the ones with the greatest likelihood of injury or death.
That's not a criticism of google or the technology. It's just a recognition that as in everything, you have to make choices.
All of that is non-sequitur to the fact that their rates are not based on good math. The accountants don't care how you determine your price. They just care that the amounts coming in and going out are what has been declared as the amounts that should be coming in and going out.
Almost. The independent auditor definitely doesn't care what the amounts are but does care that they are calculated and recorded appropriately. Insurance companies (and any company that is self-insured) has to post a liability for the unfunded part. So, if the actuarial reports show that you have an IBNR (incurred but not reported) liability and claims in process and projected claims that exceed the reserves created through your funding, you still have to post the liability and associated loss. That loss and associated liability lowers shareholders equity and if it is large enough could make the insurance company (or self-insured company) balance sheet insolvent where they have more liabilities than they have assets. If that occurs, the state could step in because the insurance company can cover its obligations or if a self-insured company, creditors can force the company into bankruptcy.
As such, the accounts and auditors, being independent, don't care what premiums the company charges, however, the CFO and BOD do. Sure, policies need to be priced to be competitive in the market, but not so much that they lose money or shareholder equity per contract written. Corporate officers can be held liable for not exercising due diligence including pricing policies for contracts.
So, it is safe to say that if contracts are being priced without rhyme or reason just because of what some other insurer is pricing them then either a) there is enough excess profit in the contracts that it is not an issue, or b) the risk calculations from the other insurer for the locale being covered are similar enough for the company in question. Anything else would deplete company assets.
Why should the Pope being insulted have anything to do with whether the earth moves around the sun? Why are you making ad hominem attacks against Galileo, and throwing out your own "evidence-free" assertions that he made "evidence-free" assertions? What does someone thinking someone else is an asshole have anything to do with their actual science?
Those aren't ad hominem attacks, the whole Galilelo thing is extremely well documented. I'll grant you the asshole comments, that part is just the original poster's opinion, but even that is pretty well accepted about Galilelo. But, Galileo did refer to Pope Paul V by a derogetory term in one of his remarks, that's recorded in the transcript.
The irony is that the whole thing was not about science. The Catholic Church already accepted the theory of heliocentricsm. What they held was that Galileo's proofs did not actually proof it based on the aristotlelian system that was standard for the day. Galileo went to the church to settle the dispute between him and the scientific community that held the same position. He thought that since the Jesuits agreed with the heliocentric model that the pope would too. But that question wasn't the questioned posed. Instead what the tribunal was asked to decide, by Galileo himself, was whether or not his proof met the requirements to be put forth as a truth instead of a theory. The Church said no.
So, I don't know if Galileo was an asshole, but based on the transcripts he was pretty arrogant. And when he didn't get the answer he wanted the second time, he basically told the pope to go screw himself. The rest is history.
He was also correct about heliocentricism but wrong as to why as he based a large portion on Copernicus' theory for the non observance of parallax shifts of starts which modern science has shown to be wrong (actually as soon as better equipment was available, Copernicus was shown to be wrong in this area). So what is it called when you get the right answer but for the wrong reasons?
Galileo was friends with Urban the VIII, but it was Paul the V that the controversy began (Urban suceeded Paul). Actually, the controversy was not about the actual science but more politics. There were already theories of heliocentricity from both Copernicus and Kepler who preceded Galileo. The standard for science back then was based on an Aristotlian system. In proving his work, Galileo relied on Copernicus, and while heliocentricity was more or less accepted in the scientific community and many in the Catholic Church, there was much dispute about the great distances between starts that Copernicus theorized to deal with the abscense of parallax shifts. The problem for Copernicus was one of crude instruments, but because his theories were not universally accepted by the astronomical community of the time, Galileo, basing his proof on Copernicus failed the Aristotlian rigor needed to for proof.
Galileo disagreed and took it to the Church assuming that since the Jesuits agreed with him, the Pope would, too. But the Pope sided with general astronomers of the day and said that he was free to teach his theory but it was not a proven fact. Luther basically said the same thing to Kepler 10 years earlier, but the Lutherans don't get in trouble for it.
Even after Galileo was placed under house arrest for publishing his work as fact instead of theory (which is what the dispute was about), the Church provided housing for him, built him an observatory, fed him, provided servents for him, paid him to do research and a host of other things. It was probably the most comfortable house arrest in history.
Anyway, there were large egos involved and Galileo refused to change his position and said that he was correct and the Church was wrong. While history has shown his theory to be correct, it is for the wrong reason. Copernicus was wrong on the parallax shifts and if they had better instrumentation he would have seen the shifts. So in a way both sides were correct at the time. The heliocentric model was correct, although that was never really disputed, but the Church was correct in that it failed the rigors of scientific proof.
Only in your head is 20 years off a "rush to production." But yes, many of the problems will take many years to solve, however, not the FUD you're throwing out there. Those "problems" are in some cases actually benefits, others of your "problems" have already been solved and the rest can be solved by half the posters to slashdot if they are not afraid of trying (which obviously you are). What's next, are you going to complain that the animatronic driver is not lifelike enough?
You know, since you brought up the whole man on the moon thing. One only need to look at NASA's AS-204 (Apollo 1) tragedy to see what rushing designs can do. Nobody is saying to sit on these cars for 20 years, but by Google's estimate, it will take 20 years before there are enough of such vehicles on the road to effectively mitigate the risk from drivers not driving these cars. Don't like the 20 year number, go talk to Google. They are actually pretty forthright with the capabilities and limitations of the vehicles. They are trying to protect themselves from the over-hype that is flying around about these vehicles.
And if, as you believe, that 1/2 the posters to slashdot could solve the remaining problems Google is facing before going live, well, maybe Google should hang out here more often and pick their brains. One would think, though, that their engineers are better equipped to deal with the issues than the average slashdotter. But who knows, maybe you are right. If so, it's time to unload some Google shares.
Jeez, Joe. How many dumb excuses are you going to manufacture in a feeble attempt to discredit self-driving cars? Just about every circumstance you mention happens today with regular taxis and current drivers. Do you think NYC taxi drivers today get 100% fidelity in understanding the destinations of their passengers? It's far more likely that the driver himself will have an accent, but even then you are not accounting for the millions of passengers who speak little or no English. The reality is that even with real taxi drivers the destination will increasingly be communicated non-verbally. You punch in where you want to go on your smart phone and the car picks you up, drops you off and you get charged automatically. No struggling to understand which Hyatt you mean, or who gets which cab.
For whatever reason, you don't like change, but please try a little harder to come up with some plausible problems that don't exist today. Just about every post you've made on this thread is filled with faulty assumptions and bad math.
I'm not sure why you think my name is Joe, probably from my nickname, but that is just from a persona from something unrelated. However, I actually embrace change. Particularly, I embrace change in technology fields. However, as any other VC will tell you, it's not all about the technology, but the perception of the technology by the public.
For instance, it is well accepted that betamax was better technology than vhs, but for a number of reasons the best technology did not win. Likewise, when technology is overhyped in what it can or will do, that may do great for initial sales, but it hurts investors, particularly VCs. That's why, if you look closely, Google and Google's engineers aren't hyping up the cars.
Yes, they are a great piece of technology and they have the potential to change transportation. But they don't claim they will be accident free. They don't claim that they will solve all of these various problems that you see people claiming and they even admit they will create new problems. Now why would they do all that? Because they know for the cars to be successful, they need to avoid the hype. The last transportation vehicle that was overhyped was promoted as unsinkable and it sank. Google does not promote the cars as uncrashable. Their own research shows that statistically their cars will be involved in accidents. Chaos theory guarantees it. No, what Google promises is that their vehicles will be less likely to be involved in an accident than a human driven vehicle. There data also supports that but that is a far cry from what people are claiming. They do not claim it will be the safest vehicle. They do not claim it won't prevent every kind of accident. The do not claim it will bring world peace and feed the hungry and everything else. They simply claim that it will be less likely to be involved in an accident than a human driven vehicle. That in itself is still pretty impressive.
All of the rest of the discussion, and you can believe me or not, but they have been going on with Google and VC and potential commercial buyers are involved with the public perception. That is where things like language barriers or using smart phones or body fluids come from. They are all mentioned in Google's investor literature and are in areas that they are working on addressing.
Google is releasing these first vehicles in the commercial market instead of the consumer market. Why? Because it is easier to convince a CEO or CFO of the value in practical terms instead of all of the hype. The technology pretty much speaks for itself. On the otherhand, what commercial accounts will be interested in is the impact to the bottom line and the ROI.
Things like having a car pick you up after keying something in on your phone isn't going to happen anytime soon. Most states don't even allow google's car and those that do require a driver behind the wheel. Eventually those laws will change, but states will be slow to change until the technology has proven
Better watch it, as I have found out on this thread, posting anything even remotely critical of Google's vehicle is a sure way to get modded down. Only time will tell how safe Google's vehicles are. I have a good idea, but several NDAs keep me from elaborating. I will point this out. The goal of Google's cars were never to be accident free. That was determined early on to be cost prohibitive. No, Google only promises that it is safer than human drivers, which is a much lower standard to beat.
This is smart of Google. The last time a vehicle was touted by a manufacture as the safest, the word used was unsinkable and that was a titanic mistake. Google doesn't want to make that mistake. They know it is impossible to account for every contingency and it is likely that the vehicles, once in wide spread use, will be involved in accidents (although most likely they will not be the cause of them). Google's engineers know that even the best AI can't see around blind curves or break the laws of physics to avoid something unexpected. At best it will do better than a human at worst it will do no worse.
If you want accident free driving, it's not going to happen. Chaos theory would indicate that it is impossible. However, if you want self driving cars with an acceptable level of risk (whatever that level may be), then that is possible. That is what Google is developing.
These problems are only solvable if you bother to spend a minute to think about them. Who has time for that? I'll believe self-driving taxis can work the day they're finally able to put a man on the moon.
Of course one can rush to production and skip all of the analysis you seem so glib about and spend your time in court with a bunch of lawsuits or you can spend the time up front analysing what could go wrong and come up with ways to mitigate the risk. The choice is totally yours.
Hmmm, they must have changed the design, but unless it has 360 vision, the most common backing accident with a child is with the child along side the vehicle and then somehow getting under the wheel, whether by reaching for something or falling or reaching for the moving vehicle or any number of scenarios. I guess an intelligent car can just sit and do nothing while a child is within 10 feet of it or turn over control to the manual driver.
No, they are not. I used to write insurance rating software. You know those companies that would offer to compare rates against different companies to get you the lowest price? In California, we were the major supplier of the software that does that. With that, we had very close contact with all of the insurance companies, as their agents would use our software to quote rates.
Contrary to 'common knowledge', auto insurance policies are not well thought out mathematical calculations. They are kludge on top of gut feeling on top of assumption on top of more layers of kludge, gut feeling and assumption. What a large number of the companies do is just clone a competitor's plan, and than change calculations based on what they assume will give them good results. I can't count the number of times I called a company to get clarification on their calculations and the response I got was "I don't know. How are you calculating it? We will just calculate it the same way you do.".
That's interesting because when auditing insurance companies, regardless of what premiums they charge they need to come up with the IBNR and various other claims data and they have to be able to show that their premiums charged and reserves on hand are enough to satisfy the various statutory and accounting requirements. I have no doubt that there is pricing pressure, but, from the actuary and accounting side, the marketing pressure is ignored, unless they are willing to take a qualified audit opinion, which most insurance companies are not. But maybe California rules are different than other states, but that still wouldn't explain how they get around GAAP and the auditing standards.
Your amount really only includes worst case scenario. Plenty of accidents happen where the damage is less than $25,000. I would say the vast majority of people have cars that are worth less than $10,000. Also, $300,000 for liability might be right in some cases, but most of the time that won't happen. I guess in the US it's more likely, as they don't have universal health care, but I still don't see the average payout being anywhere over $10,000. Many accidents will simply go unreported because people don't want their premiums to go up, or get demerit points on their license.
Your insurance company is insuring against the risk associated with whatever you policy states, so if you have a vehicle worth 25,000 and 300,000 in liability, that is the risk they calculate your premium on. If that premium is $900/yr, then the risk of them having to pay out is very low, which is the point I was trying to make.
Your bad assumption is that the vast majority of accidents require paying out the full liability and vehicle coverage. They don't.
Traffic statistics, btw, are 3.1 billion vehicle miles driven per year and around 10 million accidents. Or in other words one accident every 300k miles.
However, that assumes that people only have one accident and yet a few posts up is a person who has had 6 in 300K miles. You can't calculate accident rates that way. My bad assumption is based on the actuarial reports of our own self insured auto fleet. Yes, I over simplified it, but, basically, of the 100 drivers we have statistically less than 10% of them will be in an accident in any given year. Our drivers average about 80,000/yr. So in one year, they drive the equivalent of what the average person does in four. Statistically we would expect 10 drivers to be involved with accidents of some sort this year. Accidents even include another vehicle backing into our parked vehicle (something likely to be the case with a google car, too, and actually the number one reported accident). In addition, the 10% rate is for being involved in any accident. The study shows the same 100 drivers being involved in an accident that they were at fault in is less than 2%.
A 10%/yr accident rate does not mean that in 10 years all 100 drivers will be in an accident any more than a 50/50 chance of having a boy means that every family with two children will have a boy and a girl. Each probability stands on its own for each new year (or birth in the boy/girl example). Some routes are more prone to accidents then others as are some drivers. We have drivers with over 10yrs experience and 1M miles without a single accident.
My point is that you can not evaluate risk or mitigate it with something as simple as taking the number of accidents and dividing by the number of people. That will only yield meaningless numbers. The insurance industry has a whole "science" built around actuarial studies to understand the risk involved with insuring drivers and how to mitigate that risk. While that process is much to complex to deal with adequately in /., it should be enough to state that a handful of google cars not having any accidents is no more relevant risk wise than a handful of Ford Focus' not having any accidents.
Google cars probably are as safe as people are implying they are, but it would be better if there was industry standard data to support that claim. Lack of such data means there is just anecdotal data and that really isn't data at all.
You think a company like Google, that sends cars down every road taking pictures every 10n feet or so where n is small, is relying on your maps, do you?
I don't.
Google says it uses government sources for their maps and traffic data, so, while probably not the paper maps, it probably is the same government GIS data that the paper maps are generated from.
Not that this will stop anyone the first time the car backs over a kid, despite their excellent safety record.
The Google cars have backup cameras, radar, and bump sensors. They have been specifically designed and tested to not run over kids/pets while backing, under many different light and weather conditions. So your scenario is very unlikely to happen.
A much more likely scenario: After self driving cars are common, some human driver backs over a kid, and people ask why we should continue to allow humans to drive.
By the time a bump sensor responds it is too late. As for backup cameras, radar, etc. It is all a matter of physics and how fast the car can decelerate versus how fast the object is moving that just came into its field of so called awareness. Chances are, backing out of a driveway will be slow enough that unless the kid is on something moving quickly, it will stop in time. On the other hand, most often, children aren't hit directly from behind but glance off the rear fender and fall beneath the wheel. It is hard to see how a google car will prevent that.
You simply cannot calculate the chance of having an accident that way. Driving 300,000 miles in Montana is going to be different than driving 300,000 miles in New York City. If you want accident statistic rates, one need look no further that auto insurance. If it was a matter of dividing total mileage by number of accidents, none of us could afford our premiums.
If your articles don't make sense. The first one, the pro google article (what would you expect them to tell you about the safety) claims that people have 1 accident per 165,000 miles driven (not taking gender into question). The second article, even if men drove the same as women, would be 1 accident every 175,000 miles. Averaging the 2 rates given yields a rate of 1 accident approximately every 186K miles and in reality the number would be higher as there are more male drivers than female drivers. But even at the 186K figure that is 13% greater than the first article's number. Now, it has been a long time since I studied statistics, but a 13% variance in something that is measuring a finite set seems that there is bad data somewhere.
In reality, it goes back to Mark Twain's there lies, damn lies and statistics. Without valid samples, statistics are pretty much meaningless. Without knowing the types of accidents, any projections on deaths are meaningless (how many of those accidents are in parking lots or while parallel parking?). If your goal is to reduce fatal accidents, you can do that now, by lowering speed limits. That would be more effective now versus 20 years from now, the earliest projection when google cars will reach enough percentage of vehicles on the road to impact highway safety).
Until then, all we know is the project accident rate for a google car. Your insurance company has that same figure for you, too and they figure your premium on it. If you premium is manageable, then you are not an insurance risk and you are no more likely to be in an accident than a google car. If it is high, then you are. Plain and simple. Insurance companies are in business to make a profit and they are pretty good at analyzing the risk involved with insuring drivers.
I am pretty sure the average person has an at fault accident more often than 300k miles driven.
I'm pretty sure you are wrong or the average person's car insurance would be significantly higher than it is today. Assuming you are correct, more than one would equate to at least 2, so that would be at least 1 at fault accident every 150,000 miles. Full coverage with 300,000 liability is around $900/year assuming you are over thirty in this part of the country. Driving 15,000 miles/yr would mean 1 accident per 10 years, so the insurance company will collect 10,000 from you in premiums and potentially pay out $325,000 in claims (300,000 liability and 25,000 vehicle). If the average person is so likely to have an at fault accident, how would the insurance company stay in business in that business model (which was intentionally over simplified for this example)?
No, the fact of the matter is that most people never, ever are involved in an accident at all. That is why insurance, while not cheap, is relatively so compared to the cost being insured. Think of it like life insurance. A 20 yr old can get some pretty cheap life insurance, why? because the likelihood of the insurance company paying is very small. The same can not be said for an 80 yr old.
Agreed, 300,000 miles without an accident isn't that awesome.
Yes it is! Many of those miles are not cruising the freeway, but on a test track under conditions that were designed to cause an accident. Test dummies have been used to simulate pedestrians stepping into traffic. Other cars pull in front, or cut off the Google car, or drop objects onto the road. The self driving cars have been able to avoid thousands of accidents where a human would likely not have been able to react in time.
So you are saying that the 300,000 miles is in a laboratory controlled environment instead of a real world environment? Boy that instills confidence, because we all know how realistic that is. I wonder how many Boeing 787 had battery fires in their controlled test environments instead of real world environment? I would hazard a guess of not too many or the problem would have been addressed before releasing the plane. Controlled tests can only go so far. Obviously if you fail the controlled ones you will fail the real world ones, but passing controlled tests doesn't mean it will pass real world tests.
Hopefully, you are just posting without knowing and the google researchers actually did most of their testing in the real world and not on some test track. Otherwise, the vehicles shouldn't be allowed until proven safe.
Google's definition of "accident":
Noun
1. An unfortunate incident that happens unexpectedly and unintentionally, typically resulting in damage or injury.
2. A crash involving road or other vehicles, typically one that causes serious damage or injury.
Accidents can also be caused by chance, but the word itself doesn't have to mean that. When someone says there was an accident somewhere, they aren't (necessarily) implying that nobody was to blame.
I think replacing human driven cars with these things would save a lot of lives.
In risk management there is a big difference between incident and accident. When two airplanes fly too close (what is call a near-miss), that is an incident. If they actually hit, that is an accident. All accidents are incidents, but not all incidents are accidents. What is needed to evaluate the google car is the incident rate, not the accident rate. Why? To minimize accidents, you need to minimize incidents. If google cars are involved in a high rate of incidents, even if they avoid accidents, then the risk of an accident is high.
Think of it this way. Most teenagers do not have accidents, but they do have incidents. Accidents always occur from incidents, so insurance premiums are higher on teenage drivers. It is not the accident rate that is important in evaluating the self driving cars, it is the incident rate. Because even with low accidents, if there are high incidents eventually there will be high accidents.
The google car already has over 300k miles on it without a single at-fault incident. Although I thought the law required a person to be in the car ready to assume control at all times?
I drive an original 1973 VW Beetle every day with over 300K miles on it without a single accident (at-fault or otherwise). I would not use that statistic to say that all VW Beetles are safe vehicles. Just because a small handful of these cars have been tested does not mean that they are safe in the real world. To do a proper study, you have to have a large enough sample size. Maybe that has been accomplished, because that probably only needs around 1,000 vehicles. However, testing those 1,000 vehicles in different locales and conditions would seem to require more than 300K miles to have a statistically valid sample size.
I'm not knocking what google has accomplished, but really, has the car been tested in enough scenarios that it is safe or is it safe enough that the cost of making it safe outweighs the cost of the likelihood of injury (kind of like Ford and the Pinto)?
They will be far less likely to back over a kid, or confuse pedals like oldsters around here love to do. This is because the outside of the car can be covered in sensors instead of being a hinderance to visibility to the driver.
Really? statistically, what is the likelihood of a taxi backing over a kid or even being driven by a senior citizen. These are self driving cars for consumers, these are commercial vehicles such as taxis and delivery vehicles.
As for covering a vehicle in sensors instead of being a hindrance to visibility to the driver, the same visibility requirements exist for human driven vehicles and self-driven vehicles because humans have to be able to drive self-driven vehicles on occasion, so if you need to cover all the windows with sensors, it isn't going to work.
If your main concern is backing over things, they've had dash board cams and backup sensors for years now.
Lots of things. And they will.
But statistically, it'll probably be better than having humans behind the wheel. Not that this will stop anyone the first time the car backs over a kid, despite their excellent safety record.
Think about that. These are going to be used as taxis in New York. So, not only will you have to be able to get one, but it will have to figure out where you are going, whether you speak with an accent or not. Of course, you could just type your destination into your smart phone and forgo the talking, but if the car can't get the voice recognition algorithm correct, then that makes the safe driving algorithm suspect.
Then who is going stop all sort of stuff from happening in the back seat? Will the next person really want to sit in other's bodily fluid? Or what if two people want the same cab? Will the car choose or will they fight it out and winner take all?
It seems that self-driving cars have fewer complications for individual drivers instead of as a fleet of taxis, but then, you could always still keep a taxi driver in them defeating the point of having them in the first place.