sort of OT autonomous car fatality

[MattTuck]

Originally Posted by Kontact

As cyclists, we absolutely do. The vicitim was hit on the right side of the far right lane by a vehicle that should have seen her and gone around. If she was riding in the direction of traffic at 10 mph on the right of the right lane she would be just as dead.

What do you do when you ride to make sure that overtaking motorists observe you and don't run you over from behind? We ride on faith that vehicles won't simply ram into us from behind because they are blindly staying in the lane markings.

Well, for one, I don't typically ride at night. And when I do, I have several old lights that I would mount, and wear a reflective vest. In fact, I just got a new light for day time use. Two, I turn around and look at cars to let them know that I see them, and that also allows me to see if they are traveling fast, giving me a wide clearance, etc.
I have ridden with a mirror in the past, but I found that relying on that made some drivers feel like they could come closer to me, and I prefer to look backwards now even if I have a mirror. I think that when a driver sees that you know they're there, their behavior improves.

In my view, there is no way to eliminate all risk from these environments. You're dealing with heavy high velocity objects with huge amounts of kinetic energy, and potentially unpredictable behaviors due to the operators. The best you can do is take actions to minimize the risks you think are the most dangerous.

[Kontact]

Originally Posted by MattTuck

Well, for one, I don't typically ride at night. And when I do, I have several old lights that I would mount, and wear a reflective vest. In fact, I just got a new light for day time use. Two, I turn around and look at cars to let them know that I see them, and that also allows me to see if they are traveling fast, giving me a wide clearance, etc.
I have ridden with a mirror in the past, but I found that relying on that made some drivers feel like they could come closer to me, and I prefer to look backwards now even if I have a mirror. I think that when a driver sees that you know they're there, their behavior improves.

In my view, there is no way to eliminate all risk from these environments. You're dealing with heavy high velocity objects with huge amounts of kinetic energy, and potentially unpredictable behaviors due to the operators. The best you can do is take actions to minimize the risks you think are the most dangerous.

I am taking the point of view that a machine that sees in lidar, radar and amplified video can see at night, so this isn't a failure to be seen but a failure to take action. The Uber never reacted to the pedestrian and struck her at full speed centered in its lane.


We act on faith that cars aren't going to jerk the wheel into us when we ride, walk or drive around other cars. We can game that to an extent, but the closing rates of vehicles is beyond see and avoid reaction times. Especially since bicycles require a countersteer to make sudden maneuvers - we have slow emergency maneuver speed.

[alancw3]

arizona police have released the video of the accident both inside and outside of the vehicle:

https://www.youtube.com/watch?v=hthyTh_fopo

looking at the video i am actually surprised that the autonomous car did not stop or even attempt to slow down. perhaps i am excepting to much from self driving! i certainly would have attempted to stop as a driver. just saying.

[tuscanyswe]

Edit: Its so terrible to watch stuff like this. Just sad all way around regardless of fault

[merckxman]

Scientific American article about the accident:
https://www.scientificamerican.com/a...-blind-spots1/

[alancw3]

Originally Posted by merckxman

Scientific American article about the accident:
https://www.scientificamerican.com/a...-blind-spots1/

interesting article. looks like self driving cars still have a ways to go to truly be autonomous.

[Kontact]

Originally Posted by merckxman

Scientific American article about the accident:
https://www.scientificamerican.com/a...-blind-spots1/

The article has "blind spots" in the title, but then goes on to say that there is no good reason the layered senses of the car didn't see the victim.


I'll be very interested when they tell whether the victim was "visible", whether she was recognized as an object in the road and whether the system attempted to make a control input.

[ripvanrando]

The real challenge, says Pratt, a former academic and government program manager in robotics and intelligent systems, is developing a vehicle that can drive in "very difficult domains," such as rainy weather or crowded roads. That's level five, and Shladover, for one, says he wouldn't be surprised if it's 2075 before we get there

Shladover believes AV companies need to be much clearer about the "operational design" of their vehicles—in other words, the specific set of conditions under which the cars can function without a driver's assistance. "But most of the time they won't say, or they don't even know themselves," he says.

Congress is now considering legislation that would allow AV-makers to deploy the cars so long as they are deemed as safe as current vehicles. Even that is a high bar, Shladover notes

How much less safe than a driver operated vehicle is good enough?

http://www.sciencemag.org/news/2017/...riverless-cars

[72gmc]

I've been quietly hoping that non-human sensors and automated responses might increase cyclist safety. That we could add something to our bikes or clothes that is sensed by vehicle systems much more reliably than we are sensed by humans who are staring at their phones.

This Uber incident doesn't necessarily eliminate that hope, it just tells me they aren't there yet. And it tells me Uber doesn't have an entirely professional approach to this effort, which is not surprising given that it's Uber.

[Climb01742]

There's a profound question at the heart of this tragic situation:

We humans are imperfect (like when we drive). Technology offers the possibility of correcting some of our flaws (like when we drive). Getting there will have its price and tragic consequences.

No one should die because any company pushes technology forward too fast, or too sloppily. But equally, people should not continue to die if, someday, technology could prevent it. (Think of medicine. Medicine's path to progress costs lives, but now it saves lives.)

Staying where we are isn't a great option. Pushing technology forward too fast isn't a great option. Question is, how do we progress humanely? If anything good could come out of the awful situation it might be trying to find that answer.

As an aside, it's hard not to see some of these same issues involved with Facebook. When profit is your only goal, technology can f' over a lot of people and potentially democracy. Who's watching the technology? Not the folks who make it, that's for sure.

[OldCrank]

I just saw that video-
anybody else think it looks ummm... enhanced?

In the critical moment just before she was visible, the woman's reflective sneaker looked like it was sliding towards the car, Very Fast.

And then, there she was- totally lit up. Very focused streetlights?
Headlights that stopped at 30' from the car?

Curious.

Also I feel bad for that paid rider. Instant pariah.

Finally, she was Walking the bike. A pedestrian.
Just a legal nit.

[ripvanrando]

Originally Posted by Climb01742

There's a profound question at the heart of this tragic situation:


No one should die because any company pushes technology forward too fast, or too sloppily. But equally, people should not continue to die if, someday, technology could prevent it. (Think of medicine. Medicine's path to progress costs lives, but now it saves lives.)

Staying where we are isn't a great option. Pushing technology forward too fast isn't a great option. Question is, how do we progress humanely? If anything good could come out of the awful situation it might be trying to find that answer.

I thought of medical device development and pharmaceutical science and trials, too. I spent 30+ years in this field and am considered an expert in design and development. Aside from regulatory controls during design and development, there are phases trials to minimize risks and unintended consequences for medicinal products. The risks and benefits are carefully evaluated in a step by step fashion. What did the development process and risk analysis look like for Uber's AD car? What were the test cases before they made Tempe population their guinea pigs? What approval process did they have to go thru with any government agency to take this product from the lab to the streets? Who decided the risks were acceptable? How? Upon what data?

This failure is not even at the ethical level. What would the AD vehicle do if being left crossed by a truck and there was a bicycle to the right? Take the left cross or kill the bicyclist? What if the oncoming left cross was a bicycle,

[FlashUNC]

The video is galling. The radar/lasers and all the other bells and whistles are supposed to avoid this kind of thing. That's essentially the entirety of why we're being sold on this kind of autonomous car future, so stuff like this doesn't happen.

I can understand this if this were a human driver given all the limitations of our senses. But it ain't like radar cares how dark it is out.

[MattTuck]

Originally Posted by ripvanrando

I thought of medical device development and pharmaceutical science and trials, too. I spent 30+ years in this field and am considered an expert in design and development. Aside from regulatory controls during design and development, there are phases trials to minimize risks and unintended consequences for medicinal products. The risks and benefits are carefully evaluated in a step by step fashion. What did the development process and risk analysis look like for Uber's AD car? What were the test cases before they made Tempe population their guinea pigs? What approval process did they have to go thru with any government agency to take this product from the lab to the streets? Who decided the risks were acceptable? How? Upon what data?

This failure is not even at the ethical level. What would the AD vehicle do if being left crossed by a truck and there was a bicycle to the right? Take the left cross or kill the bicyclist? What if the oncoming left cross was a bicycle,

Yeah, I'm with you on most of this. Medicine has this idea of informed consent. That, hey, something might be potentially dangerous, but if you understand the risks, we consider it fair because there may be some benefit to you or society.

This woman is essentially the victim of a 21st century Tuskeegee Study. That said, you can only test this technology so much in the lab before it needs real life experience to get better, and in that context, accidents are sure to happen. Even with more regulation, there is no guarantee that this stuff will work -- and even if it does work in general and make the streets safer than with human drivers, they may not drive accidents to zero... so there will still be costs.

That's why I think the best thing other than clear communication is a regulatory structure that makes companies pay up front for the eventual cost that is born by the victims. For each mile of road that your autonomous vehicle drives, it is $x (not sure of the exact amount) would be one approach. Another would be a flat fee for a license to test an autonomous vehicle within your city. I prefer the first, because it makes companies carefully consider how they will deploy, monitor and examine their test data. If you allow them to costlessly roam the city, each incremental mile matters very little, and so you have much more lax safety measures.

[Kontact]

Originally Posted by MattTuck

Yeah, I'm with you on most of this. Medicine has this idea of informed consent. That, hey, something might be potentially dangerous, but if you understand the risks, we consider it fair because there may be some benefit to you or society.

This woman is essentially the victim of a 21st century Tuskeegee Study. That said, you can only test this technology so much in the lab before it needs real life experience to get better, and in that context, accidents are sure to happen. Even with more regulation, there is no guarantee that this stuff will work -- and even if it does work in general and make the streets safer than with human drivers, they may not drive accidents to zero... so there will still be costs.

That's why I think the best thing other than clear communication is a regulatory structure that makes companies pay up front for the eventual cost that is born by the victims. For each mile of road that your autonomous vehicle drives, it is $x (not sure of the exact amount) would be one approach. Another would be a flat fee for a license to test an autonomous vehicle within your city. I prefer the first, because it makes companies carefully consider how they will deploy, monitor and examine their test data. If you allow them to costlessly roam the city, each incremental mile matters very little, and so you have much more lax safety measures.

Maybe they should be street testing the technology with vehicles built out of pillows with bright flashing lights rather than 5000 pound Volvos.