3 ms·

Nice video, but the evidence is not clear that was human or computer error. It would be nice to make that distinction somewhere..

symlinkk 10y ago
There is a red light. The car, which is being driven by a computer, did not stop. That is a computer error.
- jgacook 10y ago
  The Uber self-driving cars have a human in the driver's seat that is able to take control of the vehicle. It's not clear whether the computer or the human was in control when the car blew the red light.
  - unclebucknasty 10y ago
    Some variation of that comment has been oft-repeated on this thread and it's exactly what Uber wants. One of the main functions of that human is to accept blame (or at least obfuscate culpability) and absolve the tech in the event of an accident. There are a ton of scenarios wherein it would be virtually impossible for a human--especially one lulled into inaction and passivity for most of the ride--would be able to intervene quickly enough to avert an accident once it became apparent that the tech was failing.
    - jodrellblank 10y ago
      And you're absolving the legally responsible human, so you can blame the implied-Evil corporation, because... why? Humans who are lulled into passivity by highway driving don't get to escape the blame if they crash because they couldn't respond quickly enough if the situation changes suddenly, regardless of 'tech failing'. If you have SatNav and it says 'drive down here' and you do, and it's a one-way street, you don't say "that's exactly what Garmin want, a human to absolve them when their tech fails". Because the driver is in the position of responsibility, the tech is not.
      - unclebucknasty 10y ago
        >so you can blame the implied-Evil corporation, because... why? Because the "implied-Evil corporation" offers technology which claims to be capable of doing something, entreats you to confer upon it some degree of trust, then holds you responsible when it fails. Because lives can be lost as a result of such failures and the company's stance represents a cynical and cavalier regard for this fact. Because, they want the cachet (and the cash) without the responsibility. >Humans who are lulled into passivity by highway driving That analogy isn't even close to holding. In that case, you are understood to be in control of the vehicle. In fact, you are actively giving minor corrective inputs at all times. It's nowhere near the same as the car saying, "here, let me do that for you", but, you are supposed to sit at the ready, hands hovering slightly over the wheel and foot over the brake, ever scanning and ready to take over? That's asinine, and it's a flawed model. At a minimum, your reaction time will be slowed as you must first recognize that the car is not responding properly in, perhaps, a split second when it suddenly does something unexpected. Look, it's simple: the car needs to either be fully ready and safe, or it's not. This have-it-both-ways Tesla-invented PR spin that says "our tech is statistically safer, but it's your fault for not correcting it even if it does screw up" is a pernicious subterfuge and I personally don't like how it defines our relationship with technology.
        lisivka 10y ago
        Car is not 100% safe even with human driver. If a tech makes less mistakes than a human, then the tech is better. It's your responsibility to chose what is better: you, the tech, or a taxi driver. IMHO, every self-driving car must be equipped with a black-box to capture all information in case of an accident and special commission for investigation, to make rules to follow, so self-driven cars will be safer for others, and that is all. Self-driven system manufacturer can be punished only if these rules are not followed, by suspending a license to manufacture self-driving systems. Currently, we have almost no such rules at all. IMHO, the first rule for self-driving cars, until they will be mature, must be better communication between computer and human: as driver, I want to see how computer sees road and obstacles, and planned actions, and other variants of these actions to pickup. I want computer to project computer map to wind-shield and be presented with tree buttons: big red «stop» button, left button, and right button, so I will be able to see road situation, computer understanding of that situation, and will be able to correct computer or make a full stop.
- singlow 10y ago
  But do we know whether it was under computer control at the time?
- Gustomaximus 10y ago
  The video doesn't show if it was a person or computer driving. And Uber claim it was a human error though I'm not inclined to take that as proof. https://techcrunch.com/2016/12/14/uber-looking-into-incident-of-self-driving-car-running-a-red-light-captured-on-dashcam/ https://techcrunch.com/2016/12/14/uber-looking-into-incident...
- onion2k 10y ago
  which is being driven by a computer The point is that we don't know that. It may have been under driver control at the time. There's probably a good argument for having some sort of external indicator of whether or not a self-driving car is under computer control or human control. Maybe a small roof mounted beacon that has to be lit up or something.

symlinkk 10y ago

There is a red light. The car, which is being driven by a computer, did not stop. That is a computer error.

jgacook 10y ago
The Uber self-driving cars have a human in the driver's seat that is able to take control of the vehicle. It's not clear whether the computer or the human was in control when the car blew the red light.
- unclebucknasty 10y ago
  Some variation of that comment has been oft-repeated on this thread and it's exactly what Uber wants. One of the main functions of that human is to accept blame (or at least obfuscate culpability) and absolve the tech in the event of an accident. There are a ton of scenarios wherein it would be virtually impossible for a human--especially one lulled into inaction and passivity for most of the ride--would be able to intervene quickly enough to avert an accident once it became apparent that the tech was failing.
  - jodrellblank 10y ago
    And you're absolving the legally responsible human, so you can blame the implied-Evil corporation, because... why? Humans who are lulled into passivity by highway driving don't get to escape the blame if they crash because they couldn't respond quickly enough if the situation changes suddenly, regardless of 'tech failing'. If you have SatNav and it says 'drive down here' and you do, and it's a one-way street, you don't say "that's exactly what Garmin want, a human to absolve them when their tech fails". Because the driver is in the position of responsibility, the tech is not.
    - unclebucknasty 10y ago
      >so you can blame the implied-Evil corporation, because... why? Because the "implied-Evil corporation" offers technology which claims to be capable of doing something, entreats you to confer upon it some degree of trust, then holds you responsible when it fails. Because lives can be lost as a result of such failures and the company's stance represents a cynical and cavalier regard for this fact. Because, they want the cachet (and the cash) without the responsibility. >Humans who are lulled into passivity by highway driving That analogy isn't even close to holding. In that case, you are understood to be in control of the vehicle. In fact, you are actively giving minor corrective inputs at all times. It's nowhere near the same as the car saying, "here, let me do that for you", but, you are supposed to sit at the ready, hands hovering slightly over the wheel and foot over the brake, ever scanning and ready to take over? That's asinine, and it's a flawed model. At a minimum, your reaction time will be slowed as you must first recognize that the car is not responding properly in, perhaps, a split second when it suddenly does something unexpected. Look, it's simple: the car needs to either be fully ready and safe, or it's not. This have-it-both-ways Tesla-invented PR spin that says "our tech is statistically safer, but it's your fault for not correcting it even if it does screw up" is a pernicious subterfuge and I personally don't like how it defines our relationship with technology.
      - lisivka 10y ago
        Car is not 100% safe even with human driver. If a tech makes less mistakes than a human, then the tech is better. It's your responsibility to chose what is better: you, the tech, or a taxi driver. IMHO, every self-driving car must be equipped with a black-box to capture all information in case of an accident and special commission for investigation, to make rules to follow, so self-driven cars will be safer for others, and that is all. Self-driven system manufacturer can be punished only if these rules are not followed, by suspending a license to manufacture self-driving systems. Currently, we have almost no such rules at all. IMHO, the first rule for self-driving cars, until they will be mature, must be better communication between computer and human: as driver, I want to see how computer sees road and obstacles, and planned actions, and other variants of these actions to pickup. I want computer to project computer map to wind-shield and be presented with tree buttons: big red «stop» button, left button, and right button, so I will be able to see road situation, computer understanding of that situation, and will be able to correct computer or make a full stop.
singlow 10y ago
But do we know whether it was under computer control at the time?
Gustomaximus 10y ago
The video doesn't show if it was a person or computer driving. And Uber claim it was a human error though I'm not inclined to take that as proof. https://techcrunch.com/2016/12/14/uber-looking-into-incident-of-self-driving-car-running-a-red-light-captured-on-dashcam/ https://techcrunch.com/2016/12/14/uber-looking-into-incident...
onion2k 10y ago
which is being driven by a computer The point is that we don't know that. It may have been under driver control at the time. There's probably a good argument for having some sort of external indicator of whether or not a self-driving car is under computer control or human control. Maybe a small roof mounted beacon that has to be lit up or something.