Backup driver in first death by fully autonomous car pleads guilty(english.elpais.com)
english.elpais.com
Backup driver in first death by fully autonomous car pleads guilty
https://english.elpais.com/science-tech/2023-07-29/the-backup-driver-in-the-1st-death-by-a-fully-autonomous-car-pleads-guilty-to-endangerment.html
25 comments
It is possible that this tragic accident would have happened even if the operator paid full attention, but that is besides the merit of this case. Their job was to supervise an autonomous driving test, and they did not pay full attention at the critical moment, with no justifiable reason for this lack of attention:
"Records indicate that streaming began at 9:16 pm and ended at 9:59 pm. Based on an examination of the video captured by the driver-facing camera, Vasquez was looking down toward her right knee 166 times for a total of 6 minutes, 47 seconds during the 21 minutes, 48 seconds preceding the crash. Just prior to the crash, Vasquez was looking at her lap for 5.3 seconds; she looked up half a second before the impact." - https://en.wikipedia.org/wiki/Death_of_Elaine_Herzberg
"Records indicate that streaming began at 9:16 pm and ended at 9:59 pm. Based on an examination of the video captured by the driver-facing camera, Vasquez was looking down toward her right knee 166 times for a total of 6 minutes, 47 seconds during the 21 minutes, 48 seconds preceding the crash. Just prior to the crash, Vasquez was looking at her lap for 5.3 seconds; she looked up half a second before the impact." - https://en.wikipedia.org/wiki/Death_of_Elaine_Herzberg
> Their job was to supervise an autonomous driving test
The "driver" performed her job flawlessly, by acting exactly as any other "driver" who's been rendered redundant by fully-automated systems. This "driver" exemplified the human indifference to the task when our attention isn't demanded by complete immersion in an interactive experience.
By causing the vehicle to start operation, the driver accepted responsibility and liability, per the current laws on SDCs. Therefore, "scapegoat" is a misnomer that wrongly implies innocence or detachment from the sinful deed(s). The driver accepted her fate by signing up and settling into the left-hand front seat.
The "driver" performed her job flawlessly, by acting exactly as any other "driver" who's been rendered redundant by fully-automated systems. This "driver" exemplified the human indifference to the task when our attention isn't demanded by complete immersion in an interactive experience.
By causing the vehicle to start operation, the driver accepted responsibility and liability, per the current laws on SDCs. Therefore, "scapegoat" is a misnomer that wrongly implies innocence or detachment from the sinful deed(s). The driver accepted her fate by signing up and settling into the left-hand front seat.
By causing the vehicle to start operation, the driver accepted responsibility and liability, per the current laws on SDCs. Therefore, "scapegoat" is a misnomer that wrongly implies innocence or detachment from the sinful deed(s). The driver accepted her fate by signing up and settling into the left-hand front seat.
I call bullshit.
Why was the "SDC" even on the road?
See, refusal to deal with this question is the attitude that increasingly pisses me off about tech companies. Oh, it's totally fine for me to deploy a multi-ton vehicle system with a "backup" operator that all human factors research reasonably tells us is going to be rendered unsafe by the levels of detachment induced by too good, but not good enough" levels of automation. Heaven forbid though that the company be ultimately culpable for the hazard they create. After all, they got an "independent contractor" to transfer risk onto. Likely one that poorly the nature, implementation, mechanics, and character of the system they were to supervise.
Gotta exploit the opportunity for massive profit, while saddling someone else with all the bite of the risk, because conventional testing and sane operational practices (which includes not deploying shit you know is going to be unsafe) is toooooooooo expensive and too much to ask.
Uber, apparently made the decision to deactivate the AEB system (likely because otherwise they couldn't convince anyone sane to ride in the car because they system wasn't ready). Uber decided to undertake risk transference to an independent contractor that, lets be real, every executive there knew they weren't going to have a great success rate at getting across the real operational envelope of these vehicles. The same exec team that brought you things like their networks kill switch to frustrate LE raids in other jurisdictions, and project Greyball, a system to gaslight LE in other jurisdictions as well. If anyone was "Jesus take the wheel"-ing that night, and every night leading up to it, it was Uber.
But nah... We all sit back and accept these types of shenanigans as just "cost of doing business". It's "too important" to develop safely.
I call bullshit.
Why was the "SDC" even on the road?
See, refusal to deal with this question is the attitude that increasingly pisses me off about tech companies. Oh, it's totally fine for me to deploy a multi-ton vehicle system with a "backup" operator that all human factors research reasonably tells us is going to be rendered unsafe by the levels of detachment induced by too good, but not good enough" levels of automation. Heaven forbid though that the company be ultimately culpable for the hazard they create. After all, they got an "independent contractor" to transfer risk onto. Likely one that poorly the nature, implementation, mechanics, and character of the system they were to supervise.
Gotta exploit the opportunity for massive profit, while saddling someone else with all the bite of the risk, because conventional testing and sane operational practices (which includes not deploying shit you know is going to be unsafe) is toooooooooo expensive and too much to ask.
Uber, apparently made the decision to deactivate the AEB system (likely because otherwise they couldn't convince anyone sane to ride in the car because they system wasn't ready). Uber decided to undertake risk transference to an independent contractor that, lets be real, every executive there knew they weren't going to have a great success rate at getting across the real operational envelope of these vehicles. The same exec team that brought you things like their networks kill switch to frustrate LE raids in other jurisdictions, and project Greyball, a system to gaslight LE in other jurisdictions as well. If anyone was "Jesus take the wheel"-ing that night, and every night leading up to it, it was Uber.
But nah... We all sit back and accept these types of shenanigans as just "cost of doing business". It's "too important" to develop safely.
>Oh, it's totally fine for me to deploy a multi-ton vehicle system with a "backup" operator that all human factors research reasonably tells us is going to be rendered unsafe by the levels of detachment induced by too good, but not good enough" levels of automation.
Except in this particular case, the driver wasn't just zoned out, she was streaming Hulu. I don't think she was so lacking in free will that not having a stream on was out of her control.
>Uber, apparently made the decision to deactivate the AEB system (likely because otherwise they couldn't convince anyone sane to ride in the car because they system wasn't ready).
You're saying they disabled the AEB system, otherwise people wouldn't ride in the car? How does that make sense?
Except in this particular case, the driver wasn't just zoned out, she was streaming Hulu. I don't think she was so lacking in free will that not having a stream on was out of her control.
>Uber, apparently made the decision to deactivate the AEB system (likely because otherwise they couldn't convince anyone sane to ride in the car because they system wasn't ready).
You're saying they disabled the AEB system, otherwise people wouldn't ride in the car? How does that make sense?
>You're saying they disabled the AEB system, otherwise people wouldn't ride in the car? How does that make sense?
Their AEB technology had too high a false positive rate resulting in what a consumer would quickly coin "the perception of automated vehichle as deathtrap".
Once again, leading to "we must put something on the road now (despite it not being ready for prime time).
Their AEB technology had too high a false positive rate resulting in what a consumer would quickly coin "the perception of automated vehichle as deathtrap".
Once again, leading to "we must put something on the road now (despite it not being ready for prime time).
I'm inclined to disagree here. Their role here is akin to a driving instructor. Regardless of whether the primary driver is a human student or a AI, their job was to supervise the operation of the vehicle. They did not do it and someone died. There's room to pass the buck around to be sure, but I think that was reflected in the sentencing.
Per the current laws, but what about per objective ethics? I don't care what is written in the code book.
The person shirking their duty knew the risks and was even profiting from their negligent behavior. Meanwhile a person lost their life who has nothing to gain from this. I blame the employer and the employee. Employers shouldn’t set their employees up to fail, and employees should know a risky proposition when they see it.
The stated promise of self-drive is reduction in accidents.
The accident victim didn't sign up for this, but as a member of society she did stand to gain something; to wit, harm reduction. I do see the grim irony, but, nonetheless.
Generally, it's the trolley problem. (Although in this specific case there's an element of human fault.)
The accident victim didn't sign up for this, but as a member of society she did stand to gain something; to wit, harm reduction. I do see the grim irony, but, nonetheless.
Generally, it's the trolley problem. (Although in this specific case there's an element of human fault.)
and if there is no accident, due to intervention of the driver, will the situation even be noted somewhere?
in similar threads on hn there are many near miss stories.
This assumes employees have the privilege of walking away from a risky proposition.
> Vasquez had previously spent more than four years in prison for two felony convictions — making false statements when obtaining unemployment benefits and attempted armed robbery — before starting work as an Uber driver, according to court records.
My point is that even if she had been paying full attention, the outcome wouldn't be different. If Uber hasn't wouldn't disabled AEB, it would have been different. Hence Uber is the most at fault (amongst the two of them - the pedestrian might really be the most at fault here)
> the victim only appears for a few frames before the accident.
1. It only looks that way because the camera has terrible dynamic range. If you've ever driven in real life you'd know you can see further than 100ft in front of you.
2. if for whatever reason you can't see 100ft in front of you (eg. fog), you shouldn't be driving at such a speed where you can't stop in time if an object popped into existence 100ft in front of you.
edit:
>The forward-looking Uber dashcam did not pick up Herzberg until approximately 1.4 seconds before the collision, suggesting (as the sheriff did) that the crash may have been completely unavoidable even if Vasquez hadn't been distracted in the seconds leading up to the crash.[61]
>However, night-time video shot by other motorists in the days following the crash, plus their comments, suggest that the area may have been better illuminated than the dashcam footage, viewed in isolation would suggest. This raises the possibility that Herzberg's appearing so late in the Uber video could merely be an indication that the camera had insufficient sensitivity or was otherwise poorly calibrated for the environment and setting in which it was operating. If these crowd-sourced re-creations are indeed representative of the visibility conditions on the actual night that the crash occurred, then Herzberg would have been visible to Vasquez as soon as there was a clear sight line had Vasquez only been looking ahead, refuting the assertion that the crash was unavoidable.[62]
[62]: https://arstechnica.com/cars/2018/03/police-chief-said-uber-...
1. It only looks that way because the camera has terrible dynamic range. If you've ever driven in real life you'd know you can see further than 100ft in front of you.
2. if for whatever reason you can't see 100ft in front of you (eg. fog), you shouldn't be driving at such a speed where you can't stop in time if an object popped into existence 100ft in front of you.
edit:
>The forward-looking Uber dashcam did not pick up Herzberg until approximately 1.4 seconds before the collision, suggesting (as the sheriff did) that the crash may have been completely unavoidable even if Vasquez hadn't been distracted in the seconds leading up to the crash.[61]
>However, night-time video shot by other motorists in the days following the crash, plus their comments, suggest that the area may have been better illuminated than the dashcam footage, viewed in isolation would suggest. This raises the possibility that Herzberg's appearing so late in the Uber video could merely be an indication that the camera had insufficient sensitivity or was otherwise poorly calibrated for the environment and setting in which it was operating. If these crowd-sourced re-creations are indeed representative of the visibility conditions on the actual night that the crash occurred, then Herzberg would have been visible to Vasquez as soon as there was a clear sight line had Vasquez only been looking ahead, refuting the assertion that the crash was unavoidable.[62]
[62]: https://arstechnica.com/cars/2018/03/police-chief-said-uber-...
You watched the dark video released by Uber, others went to the same street and shot videos with normal dash cams at same time of night, difference in visibility is stark.
https://arstechnica.com/cars/2018/03/police-chief-said-uber-...
https://arstechnica.com/cars/2018/03/police-chief-said-uber-...
I haven’t followed the legal contours of this case very closely so perhaps these are not interesting questions.
Who would go to jail if _no_ driver was in the car?
From what I recall, the vehicle was a Volvo and its pedestrian detection and emergency braking system was defeated by the test equipment. Is there any culpability there?
Or what about the person who planned and initiated the driving route that night, in the dark through a populated area. Are they responsible to a degree?
It doesn’t sound like the backup “driver” was engaging in engineering or testing activities, and certainly not driving, but instead was paid to be doing what a passenger might do if they thought someone else was driving.
What the driver is legally responsible for may be different from what they believed they were doing. At least, that would me my defense in the situation.
Who would go to jail if _no_ driver was in the car?
From what I recall, the vehicle was a Volvo and its pedestrian detection and emergency braking system was defeated by the test equipment. Is there any culpability there?
Or what about the person who planned and initiated the driving route that night, in the dark through a populated area. Are they responsible to a degree?
It doesn’t sound like the backup “driver” was engaging in engineering or testing activities, and certainly not driving, but instead was paid to be doing what a passenger might do if they thought someone else was driving.
What the driver is legally responsible for may be different from what they believed they were doing. At least, that would me my defense in the situation.
Apparently, the law generally indicates that the "driver" of an SDC is the one who causes it to begin operation. So, the vendors will have banks of remote operators who will theoretically bear liability and responsibility for the cars they run.
> Who would go to jail if _no_ driver was in the car?
The CEO should. If we remove the human, the liability should go to those in charge of the machine. To hell with this nonsense of a "company" absolving the humans in charge of their killings.
The CEO should. If we remove the human, the liability should go to those in charge of the machine. To hell with this nonsense of a "company" absolving the humans in charge of their killings.
what if the CEO is a new hire? because the previous one was put in jail?
This is a terrible result. It shows that companies can set up humans to take the fall for their immoral decisions. The emergency braking that Uber turned off would have saved the victim's life.
At least Uber's self driving division got the death penalty. Would have been better if the state had fined the entire corrupt enterprise into bankruptcy.
At least Uber's self driving division got the death penalty. Would have been better if the state had fined the entire corrupt enterprise into bankruptcy.
If the backup driver is guilty of failing to stop the primary driver of killing someone, then that means the primary driver is also guilty of killing someone.
I drove the same route, at the same time, at the speed limit, the days immediately following the event. I would have easily been able to see a cyclist crossing the road, with plenty of time to react.
Fully autonomous cars would work well if they could all drive at 15mph without stop signs.
Easily enough to dodge, stop, limit damage while not requiring red lights or stop and go allowing passengers and pedestrians to travel.
Easily enough to dodge, stop, limit damage while not requiring red lights or stop and go allowing passengers and pedestrians to travel.
What a bunch of fracking bullshit.
Did the driver control the speed of the car and made all decisions? Did the driver disable the automatic breaking? Could a human driving the car reasonably avoid the pedestrian?
Did the driver control the speed of the car and made all decisions? Did the driver disable the automatic breaking? Could a human driving the car reasonably avoid the pedestrian?
On the other hand, autonomous breaking systems might have actually made a difference here if they hadn't been disabled (at least in reducing the severity of the accident). And given the conditions, a human might have been driving slower. The driver here had no control over any of this, while their employer did...