Uber self-driving vehicle kills pedestrian

[gillep]

This all seems insane to me. I get that people are supportive of new tech. But seriously, how could this ever be unleashed? Maybe in a controlled environment where every vehicle is self-driving and everyone knows that every vehicle is self driving. But the risk of this happening again is surely too great for a risk averse society like ours.

I hear she wasn't in a cross walk. But should that be a death sentence? The roads are full of risk, but risk is substantially less when people in vehicles are processing their environment and will slow down or stop for a person with a momentary lapse in judgment.

These are already on the road each and every day and encounter pedestrians, bikers, other vehicles etc. every single day. I live in Silicon Valley, and I see them out all the time, especially when I go up closer to where the major tech companies are located (I live at the Southern most end of Silicon Valley, and most of the tech companies are North of me). In my experience of being on the road with them I don't believe that they are any more unsafe than human drivers, and I believe that all the data supports that as well.

 

[Mackenzie Click-Mickelson]

This all seems insane to me. I get that people are supportive of new tech. But seriously, how could this ever be unleashed? Maybe in a controlled environment where every vehicle is self-driving and everyone knows that every vehicle is self driving. But the risk of this happening again is surely too great for a risk averse society like ours.

I hear she wasn't in a cross walk. But should that be a death sentence? The roads are full of risk, but risk is substantially less when people in vehicles are processing their environment and will slow down or stop for a person with a momentary lapse in judgment.

I get what you're saying as a whole. My concern with cars like these was largely from an insurance standpoint but I did also wonder regarding technology and reliance on it (what klayfish was talking about).

At the same time in this particular situation in the OP the incident occurred at 10pm at night. Self-driving car or not seems incidents regarding pedestrians seems to increase when nighttime occurs (not a scientific statement though). If the person was just taking a light night walk, etc did they wear reflective or bright colored clothing to help increase their visibility? The pedestrian wasn't in a cross-walk. Ultimately though the driver of the self-driving car needed to maintain active vs passive awareness of their surroundings. So to me at least from the information I have I would say it's balancing more on the human error part unless the Uber vehicle had no way of taking back full control of the vehicle (I don't know the full tech Uber was using).

However, on a serious note it's quite astonishing how people still dart out in traffic to get across the street and getting hit (and killed) has happened because of that rather than wait for a cross-walk and proper signal. If the vehicle involved in the incident wasn't a self-driving vehicle we wouldn't really be having this conversation based on the actual incident. It would either be deemed the driver who failed to stop in time or abide by the cross walk (if the pedestrian was in a cross-walk) or it would be an unfortunate incident involving a pedestrian who failed to use the appropriate cross-walk.

 

[Squirlz]

Video from the car shows the woman darting out of shadows directly into the path of the car. There was no time to react.

 

[declansdad]

This all seems insane to me. I get that people are supportive of new tech. But seriously, how could this ever be unleashed? Maybe in a controlled environment where every vehicle is self-driving and everyone knows that every vehicle is self driving. But the risk of this happening again is surely too great for a risk averse society like ours.

I hear she wasn't in a cross walk. But should that be a death sentence? The roads are full of risk, but risk is substantially less when people in vehicles are processing their environment and will slow down or stop for a person with a momentary lapse in judgment.

I am fairly certain there have been pedestrians killed when they darted out into traffic in the past. Driver or driverless, it may not have mattered.

 

[gotomu212]

On your first point, not necessarily. In fact, maybe not at all. Unless you have an environment where all vehicles are autonomous, there are no pedestrians, etc...basically a 100% isolated system, it will never be as safe as it's envisioned. Having "autonomous" vehicles mixed in with semi-autonomous or fully human controlled vehicles will always create potential for major issues. While it's not impossible to have a 100% isolated autonomous environment, we're so incredibly far from it...and it has a lot of very large roadblocks in front of it. Even in autonomous situations, computers aren't perfect. So many things can still happen...software glitches (how often does an Apple product actually work as designed?), hardware failures, weather damages, etc... All it will do is shift liability focus. I don't see insurance being the driver to getting people off the wheel...at least not in our lifetime.

On your second point, yes, I'd agree. Even without autonomy, our current environment is already starting to change the model of car ownership. Sad, but true, at least IMO.

I respect your opinion on the subject and knowledge of the industry, but I think economic history just doesn’t support it. Autonomous cars won’t be risk free, but in early testing they’ve been shown exponentially safer than good drivers let alone distracted drivers. If it’s found that drivers are at fault in 80% of accidents and autonomous cars are only responsible for 20% (and that initially seems like a very high estimate of their risk) then it’s going to cost much more for someone who wants to drive since it’s shown to carry such a higher risk premium (it’s the same economics that make young new drivers or multiple accident/ticketed drivers more expensive).

As far as not in our lifetime- I conceivably have another 50 years to live if family history and current life expectancy holds....we didn’t have microwaves, communication satellites, personal computers, cell phones, the internet, even a polio vaccine 50 years ago. We just went to space for the first time 57 years ago...I can’t imagine that autonomous vehicles won’t take over driving in the same time frame.

 

[gotomu212]

But the risk of this happening again is surely too great for a risk averse society like ours.

I hear she wasn't in a cross walk. But should that be a death sentence? The roads are full of risk, but risk is substantially less when people in vehicles are processing their environment and will slow down or stop for a person with a momentary lapse in judgment.

You are right that risk goes down as drivers slow down and process their environment...the problem is that you can’t control for that. Distracted drivers, texting drivers, drunk drivers, fast drivers, inexperienced drivers, and eldearly drivers will continue to be a risk. Even a good driver can get momentarily distracted. A computer does not have those same issues (note they aren’t risk free there can still be issues like what happened).

As far as being a risk adverse society, states can’t even uniformly pass laws outlawing texting and talking on the phone while driving. We have hard evidence that this results in death. In 2014 there were 3,179 people killed by distracted drivers, (Again not apples to apples because the sample size is too low) but I’m not much comforted by a warm body behind the wheel operating it either.

 

[Klayfish]

I respect your opinion on the subject and knowledge of the industry, but I think economic history just doesn’t support it. Autonomous cars won’t be risk free, but in early testing they’ve been shown exponentially safer than good drivers let alone distracted drivers. If it’s found that drivers are at fault in 80% of accidents and autonomous cars are only responsible for 20% (and that initially seems like a very high estimate of their risk) then it’s going to cost much more for someone who wants to drive since it’s shown to carry such a higher risk premium (it’s the same economics that make young new drivers or multiple accident/ticketed drivers more expensive).

As far as not in our lifetime- I conceivably have another 50 years to live if family history and current life expectancy holds....we didn’t have microwaves, communication satellites, personal computers, cell phones, the internet, even a polio vaccine 50 years ago. We just went to space for the first time 57 years ago...I can’t imagine that autonomous vehicles won’t take over driving in the same time frame.

You are right that risk goes down as drivers slow down and process their environment...the problem is that you can’t control for that. Distracted drivers, texting drivers, drunk drivers, fast drivers, inexperienced drivers, and eldearly drivers will continue to be a risk. Even a good driver can get momentarily distracted. A computer does not have those same issues (note they aren’t risk free there can still be issues like what happened).

As far as being a risk adverse society, states can’t even uniformly pass laws outlawing texting and talking on the phone while driving. We have hard evidence that this results in death. In 2014 there were 3,179 people killed by distracted drivers, (Again not apples to apples because the sample size is too low) but I’m not much comforted by a warm body behind the wheel operating it either.

The "problem" with your argument is that you're making a lot of very, very large assumptions. In a perfect-world, utopia situation, yes computers are safer than humans. I'm talking about a situation where every single vehicle is fully autonomous, is "talking" to every other vehicle, on very controlled "roads" with an infrastructure designed to support it. Basically Jetsons' stuff come to life. But that utopian picture doesn't exist, and while I'd never say never, is unlikely to anytime soon. "Autonomous" vehicles are out there now, in limited and relatively controlled environments. They still make mistakes, they still crash...but you only hear about it when it's catastrophic, such as a fatality. When they have to interact with humans, that throws complexity into it that computer programming simply can't account for. What about the weather? They heavily rely on sensors and such. Cake them up with ice and snow, then what? They have to "read" roads. What happens when lane markings are faint, wrong or simply can't be picked up? They would need to "talk" to other autonomous cars. The list is nearly endless.

I'm certainly far, far from the foremost expert on this subject, but I definitely follow it well, I have to. Actually, Car and Driver magazine wrote a 25 page article about this topic just a few months back, very well laid out. I agree with a lot of what they said...the technology is exciting, but there are SO many obstacles and challenges that a fully automated world is still quite the question, at best. And going back to my personal tastes, just as hard as some people fight to be able to keep their guns, I'd fight just as hard to keep control over my car.

 

[PollyannaMom]

If the vehicle involved in the incident wasn't a self-driving vehicle we wouldn't really be having this conversation based on the actual incident. It would either be deemed the driver who failed to stop in time or abide by the cross walk (if the pedestrian was in a cross-walk) or it would be an unfortunate incident involving a pedestrian who failed to use the appropriate cross-walk.

I agree, especially in light of this:

Video from the car shows the woman darting out of shadows directly into the path of the car. There was no time to react.

If this had been a human driver, I really doubt they'd be charged with anything.

 

[gotomu212]

The "problem" with your argument is that you're making a lot of very, very large assumptions. In a perfect-world, utopia situation, yes computers are safer than humans. I'm talking about a situation where every single vehicle is fully autonomous, is "talking" to every other vehicle, on very controlled "roads" with an infrastructure designed to support it. Basically Jetsons' stuff come to life. But that utopian picture doesn't exist, and while I'd never say never, is unlikely to anytime soon. "Autonomous" vehicles are out there now, in limited and relatively controlled environments. They still make mistakes, they still crash...but you only hear about it when it's catastrophic, such as a fatality. When they have to interact with humans, that throws complexity into it that computer programming simply can't account for. What about the weather? They heavily rely on sensors and such. Cake them up with ice and snow, then what? They have to "read" roads. What happens when lane markings are faint, wrong or simply can't be picked up? They would need to "talk" to other autonomous cars. The list is nearly endless.

I'm certainly far, far from the foremost expert on this subject, but I definitely follow it well, I have to. Actually, Car and Driver magazine wrote a 25 page article about this topic just a few months back, very well laid out. I agree with a lot of what they said...the technology is exciting, but there are SO many obstacles and challenges that a fully automated world is still quite the question, at best. And going back to my personal tastes, just as hard as some people fight to be able to keep their guns, I'd fight just as hard to keep control over my car.

I really do hear you, and I think both of us are making a lot of assumptions. Neither of us can realistically know how the technology will continue to progress or how regulation will help or hurt the development over the next 10 years. I’m also not basing my assumptions on a risk free completely autonomous future either. I’m basing it on the tests that are being conducted over the past 5 years on the open roads with mixed use. I’m certainly not saying we will have a future without any car crashes or accidents, only that I predict it will be safer than it is now and cost factors will drive most people to convert to autonomous transportation.

We won’t convince each other today, but let’s revisit the thread 10 years from today and see which of us was closer to the reality....winner buys the other a drink (which will be delivered via a friendly self driving robot bartender hahahaha).

 

[kaytieeldr]

we didn’t have microwaves, communication satellites, personal computers, cell phones, the internet, even a polio vaccine 50 years ago.

Respectfully, yes we did. I'm well over 50, and the polio vaccine is older than I.

I knew that, but looked it up anyway. Polio is virtually eradicated. Thirty seven cases worldwide in 2016.

 

[okeydokey]

I’m sure there were many pedestrians killed the same day by cars with drivers.

 

[gotomu212]

Respectfully, yes we did. I'm well over 50, and the polio vaccine is older than I.

I knew that, but looked it up anyway. Polio is virtually eradicated. Thirty seven cases worldwide in 2016.

I stand corrected. The polio vaccine came out in 1955–I missed it by 13 years and shouldn’t have included it in my list of last 50 year advancements.

 

[EveDallas]

This all seems insane to me. I get that people are supportive of new tech. But seriously, how could this ever be unleashed? Maybe in a controlled environment where every vehicle is self-driving and everyone knows that every vehicle is self driving. But the risk of this happening again is surely too great for a risk averse society like ours.

I hear she wasn't in a cross walk. But should that be a death sentence? The roads are full of risk, but risk is substantially less when people in vehicles are processing their environment and will slow down or stop for a person with a momentary lapse in judgment.

I'm glad I'm not the only one who thinks this.

 

[EveDallas]

I respect your opinion on the subject and knowledge of the industry, but I think economic history just doesn’t support it. Autonomous cars won’t be risk free, but in early testing they’ve been shown exponentially safer than good drivers let alone distracted drivers. If it’s found that drivers are at fault in 80% of accidents and autonomous cars are only responsible for 20% (and that initially seems like a very high estimate of their risk) then it’s going to cost much more for someone who wants to drive since it’s shown to carry such a higher risk premium (it’s the same economics that make young new drivers or multiple accident/ticketed drivers more expensive).

As far as not in our lifetime- I conceivably have another 50 years to live if family history and current life expectancy holds....we didn’t have microwaves, communication satellites, personal computers, cell phones, the internet, even a polio vaccine 50 years ago. We just went to space for the first time 57 years ago...I can’t imagine that autonomous vehicles won’t take over driving in the same time frame.

Fifty years ago, people were also convinced we'd have flying cars by now.

 

[kamik86]

The "problem" with your argument is that you're making a lot of very, very large assumptions. In a perfect-world, utopia situation, yes computers are safer than humans. I'm talking about a situation where every single vehicle is fully autonomous, is "talking" to every other vehicle, on very controlled "roads" with an infrastructure designed to support it. Basically Jetsons' stuff come to life. But that utopian picture doesn't exist, and while I'd never say never, is unlikely to anytime soon. "Autonomous" vehicles are out there now, in limited and relatively controlled environments. They still make mistakes, they still crash...but you only hear about it when it's catastrophic, such as a fatality. When they have to interact with humans, that throws complexity into it that computer programming simply can't account for. What about the weather? They heavily rely on sensors and such. Cake them up with ice and snow, then what? They have to "read" roads. What happens when lane markings are faint, wrong or simply can't be picked up? They would need to "talk" to other autonomous cars. The list is nearly endless.

I'm certainly far, far from the foremost expert on this subject, but I definitely follow it well, I have to. Actually, Car and Driver magazine wrote a 25 page article about this topic just a few months back, very well laid out. I agree with a lot of what they said...the technology is exciting, but there are SO many obstacles and challenges that a fully automated world is still quite the question, at best. And going back to my personal tastes, just as hard as some people fight to be able to keep their guns, I'd fight just as hard to keep control over my car.

Some of the things that can be done with video motion detection and other sensors is pretty amazing. I'm not sure it has to be as perfect as you think it does.

I mean throw ice into the mix and a whole lot of humans don't know how to drive either. They start sliding panic and slam on the brake making it worse. A computer doesn't panic. I don't doubt that a computer could be programmed to drive on ice better then the average human.

Actually the thing I have always been most worried about with this technology comes back to the humans. It goes to how well will people make sure the cars are maintained, that becomes more important when they drive themselves and will people become reliant on the feature.

This is something that came up in my industry because we make systems that have safety features built in. There is one moveable part that caused alot of debate. It has the potential if it were to go off with someone in the wrong place it could kill someone.

There was a huge debate if there should be a sensor that disables the movement when something is too close.
One argument is that provides another layer of protection to the person looking physically before they move the item.
The other side argues that if you disable when something is there it will make the operator not look. They will rely completely on the sensor.

That attitude scares me when it comes to self driving cars because really the BEST and safest solution would be a computer to assist that can be overridden by a human paying attention.

 

[Klayfish]

Yes, sensors and such are amazing technology...but from an automotive perspective, if you want the car to be fully autonomous, they do need to be darn near perfect. Foul weather, grime, debris can effect that. That's what I was referring to when it comes to ice/snow...that can cake over a sensor and cause issues. Even clean sensor struggle. That's one of the biggest challenges cars that can partly drive themselves, like Tesla, are struggling with. The sensors have a very difficult time reading the road. That may not be the fault of the sensor. Road lane markings are faded, broken, covered by snow/ice, etc... All of that would need to be fully corrected for true autonomy. We can't even get potholes filled, let alone a complete rebuild of our infrastructure.

Yes, stability control, ABS, traction control do a far better job then humans when it comes to driving in ice and snow. But actually, because of these technologies, you SHOULD slam on the brakes if you have an issue. The notion of not slamming the brakes, pumping the pedal, etc...is from decades ago and is actually now the dangerous way to do it in a car that's equipped with modern technology. Stomp and steer is the right answer.

As I mentioned before, this is such uncharted waters. There are so many variables, unknowns, and insanely large hurdles to having a truly autonomous vehicle society. Personally, I hope it never happens. Electronic features to make humans safer drivers, I'm 110% in favor of it, but that's where it stops for me.

 

[marcyleecorgan]

When people talk about self-driving cars and auto-autonomy, I am always reminded of that lovely movie, The Fifth Element.

 

[Mackenzie Click-Mickelson]

When people talk about self-driving cars and auto-autonomy, I am always reminded of that lovely movie, The Fifth Element.

And what a great movie that is

 

[kamik86]

Yes, sensors and such are amazing technology...but from an automotive perspective, if you want the car to be fully autonomous, they do need to be darn near perfect. Foul weather, grime, debris can effect that. That's what I was referring to when it comes to ice/snow...that can cake over a sensor and cause issues. Even clean sensor struggle. That's one of the biggest challenges cars that can partly drive themselves, like Tesla, are struggling with. The sensors have a very difficult time reading the road. That may not be the fault of the sensor. Road lane markings are faded, broken, covered by snow/ice, etc... All of that would need to be fully corrected for true autonomy. We can't even get potholes filled, let alone a complete rebuild of our infrastructure.

Yes, stability control, ABS, traction control do a far better job then humans when it comes to driving in ice and snow. But actually, because of these technologies, you SHOULD slam on the brakes if you have an issue. The notion of not slamming the brakes, pumping the pedal, etc...is from decades ago and is actually now the dangerous way to do it in a car that's equipped with modern technology. Stomp and steer is the right answer.

As I mentioned before, this is such uncharted waters. There are so many variables, unknowns, and insanely large hurdles to having a truly autonomous vehicle society. Personally, I hope it never happens. Electronic features to make humans safer drivers, I'm 110% in favor of it, but that's where it stops for me.

Assuming you have ABS. My current car does but my last one didn't.

as for having to be darn near perfect... human drivers are no where near perfect... it may be viable simply to be better.

However I agree for most cases I would still like to be in charge of my car... although if I ever have to go to a city where being able to parallel park is useful a car that does that would be nice. (its so not an issue in my area I haven't done it since I took drivers ed... )

 

[gotomu212]

Fifty years ago, people were also convinced we'd have flying cars by now.

Is that really a fair comparison though? Autonomous technology is something that exists now and is being tested on the roads today. So we aren’t talking about a fantastical concept creation with few known practical applications; we’re talking about the refinement of a current technology and the time line of wide spread adoption.