AP Investigation Reveals Google Driverless Cars Have Actually Been Crashing At Twice The National Rate?

The national rate for reported "property-damage-only crashes" is about 0.3 per 100,000 miles driven, according to the National Highway Traffic Safety Administration.

Google's 11 accidents over 1.7 million miles would work out to 0.6 per 100,000, but as company officials noted, as many as 5 million minor accidents are not reported to authorities each year — so it is hard to gauge how typical this is.

I would think you could at least get the photo correct if you are going to use it and watermark it with "Auto Spies". The photo is a Google Street View mapping car and not a Driverless car.

The picture is correct. The stats were for Google driveless cars that were being tested. These include the mapping cars.

The mapping cars have drivers, so I don't see how the picture applies? The Prius in the picture is definitely not autonomous, and no where in the article did it mention a driverless mapping car. The mapping cars have $120K worth of equipment on them alone. Why would you combine both technologies at the same time at this early stage?

@terry989 - The mapping cars have been driving themselves since at least 2010

http://www.nytimes.com/2010/10/10/science/10google.html?_r=0

So you forced me to read yet another article that does not collaborate your belief that the some of the Street View Mapping cars are driveless. The only paragraph even close to this statement was:

"He said the cars did attract attention, but people seem to think they are just the next generation of the Street View cars that Google uses to take photographs and collect data for its maps."

So apparently you run with the same herd and just assume they are one in the same.

I suggest people read the article because it has some important details, such as:

Article: "...all 11 accidents were minor — "light damage, no injuries""
(This is inconsistent with the image of catastrophic damage posted on this page.

Article: "...Not once was the self-driving car the cause of the accident..."

These are important details! Miss them and you misunderstand the true meaning of the information.

"EXTRA! EXTRA! Autonomous cars are crashing all over the place!!!" This sells click.

"Autonomous cars are the safest around, but they can't make the other guy less stupid." This doesn't sell clicks.

You maybe forgetting one thing. A human driver can see and make judgements that a computer may not be able to make an accurate assessment of.

Such as weaving of a driver 4 cars ahead because he is texting. You approach with caution the while driverless car has not pieced it together.

This causes you to be in the wrong place at the wrong time and get into a collision. Since the driverless car does not have your intuition it will be in the wrong place more often over time and therefore be in more wrecks.

This is why the crash rate is higher. Now if the world was full of driverless cars then the rates would drop because the human factor would be nullified. Everything would be planned in advance.

I guess I am not entirely convince that a driverless vehicle can drive defensively.

'009 says "...You maybe forgetting one thing. A human driver can see and make judgements that a computer may not be able to make an accurate assessment of..."

What makes you believe that an "erratic driving" algorithm isn't possible? I recommend you add "Deep Learning" to your recommended reading list. It's beyond "fuzzy logic." It's a system totally unlike conventional computer programming, and it exists today. In one example of Deep Learning software in actual use, a non-medical computer research team fed the system thousands of images of healthy tissue and cancerous tissue. The Deep Learning system analyzed this info to "understand" how to differentiate cancerous tissue visually from an image, and then extrapolated (i.e., it's own learning, more that what was "taught" to the system). In a matter of hours, the system discovered several completely new ways to identify cancerous tissue visually, which were previously unknown to science! And no false positives.

The researchers (all computer guys with no medical background) released this information for free to the medical community, and now there are over 200,000 doctors (MDs) that are following what this deep learning system is learning about how to detect cancer. There's no reason why Deep Learning can't be applied to driving an autonomous car, or just about any other discipline.

There's a lot of amazing stuff going on that most people don't know about.

It is rather amazing, thanks for sharing!

Some make comparisons with how elevator technology has evolved. Initially they were entirely manual, an operator that had to shut the doors, apply the brakes, line up to the floor, and were rather dangerous machines. Eventually through technology evolution high-speed elevators and autonomous trains now proliferate the world and very rarely anyone feels that they can do a better or a safer job running them.

As enthusiasts as much as we are uncomfortable with the notion that our driving pleasures will be taken away, the majority of people are using cars as appliances and will overrule us.

Right now one the biggest issue is that Google cars can't "see" things like an open manhole cover or determine the composition of an object in the road.

If that type of perception is unavailable then what leads me to believe they have superior problems solving algorithms such as looking ahead 4 cars in traffic and making judgement to move over two lanes when passing rather than one. Yes I believe they can maintain themselves in simple traffic but it is the complex situations that worry me. Also if the road isn't mapped by Google, it can't drive on it.. so much for that new subdivision you moved into. It might be months before you can get the driverless car down your street.

Please read the following link to see where I am coming from (I'm not just pulling this stuff out of no where):

http://www.technologyreview.com/news/530276/hidden-obstacles-for-googles-self-driving-cars/

From another article in September of 2014 referencing the MIT review:

The most recent issue of MIT Technology Review has revealed a list of the things the cars can’t yet do, as confirmed by Chris Urmson, director of the Google car team. These range from the mildly problematic - for example, the cars can’t detect the nature of an obstacle, so would swerve around balls of paper as though they were rocks...

...to the downright concerning. Such as not having been tested in adverse weather conditions such as snow or rain. Or being able to detect open manhole covers or potholes.

'009 writes "...Right now one the biggest issue is that Google cars can't "see" things like an open manhole cover or determine the composition of an object in the road..."

True! I saw a TED Talk recently about people who are working on just that problem. The example they used is the scenario of a plastic bag blowing in the wind across the vehicle's path. The vehicle can easily calculate "am I likely to collide with this object," but at this time, it does not have the ability to ascertain "what does it mean if we collide?" Plastic bag: inconsequential, ignore. Rolling boulder: big bad news, take evasive action!

Smart people are aware of this issue, and they've been working on it for a while. I'm confident they'll figure it out.

Remember the days when people argued against seat-belts, claiming nobody will use them, and if they do, they'll cut you in half in an accident? Remember people argued against airbags, claiming that the explosion will knock your head right off or in the very least, make you go deaf (second claim is a genuine risk)? Remember people who argued about ABS (Antilock Brake System) claiming they can stop a car in a shorter distance than a computer can? Lots of people have been against new things that eventually became the new standard. There's nothing new about that. And I believe we're better of for many of those new standards.

BTW, I really dislike the idea of having an autonomous-only vehicle, but I realize that's just my own bias speaking. Still, that bias exists in me, and I want a drive-it-myself vehicle. But hey, I just recently dropped a lifetime of manual transmissions for my first 8-speed automatic, so that's some progress for me ;-)

009, when we see how the populace out there "drives" maybe it is better to leave that to the machines. Even if they are not perfect they are already safer than the terrible drivers out there not paying any attention. What we really object to is the mandatory imposition of rules that are alleged to keep us safe.