Self driving Uber vehicle kills pedestrian

[Minimalist]

Artificial intelligence is hopelessly inferior to human stupidity.



https://www.reuters.com/article/us-autos...SKBN1GV296

She was apparently crossing near the crosswalk but not in it.  I guess as far as Uber is concerned that makes her a legitimate target.

[Alexmahone]

How much cheaper are self-driven Ubers compared to regular ones?

[Minimalist]

It's all about not having to pay a driver.

[Anomalocaris]

On the news today.  Car was in automatic mode with a human overseer, still managed to kill a pedestrian, apparently in a crosswalk.

This ain't good.

Uber has suspended their program for now.

On the news today. Cars were in manual mode without computer overseers, still managed to kill many pedestrian, both in and out of crosswalks.

This ain’t good.

American humans have suspended their get around without walking program for now.

[Tiberius]

Self driving cars are actually very good. They have been around for a while now, this is the first case of a death.

What I really want to know is why the human driver didn’t brake. The whole reason there’s still a human behind the wheel is to act when the car fails. There’s clearly still work to do, but we also shouldn’t overreact to one case.

[Fireball]

Self driving cars are actually very good. They have been around for a while now, this is the first case of a death.

What I really want to know is why the human driver didn’t brake. The whole reason there’s still a human behind the wheel is to act when the car fails. There’s clearly still work to do, but we also shouldn’t overreact to one case.

I'm willing to bet that she stepped out from behind something (looking at her phone) and even the software didn't have time to react and/or there was not enough space between the vehicle and her to make a stop, in time. I'd also bet that the "supervisory" human in the vehicle was paying zit point shit's worth of attention (looking at their phone) to the vehicle's surroundings. After all, the car is driving itself. I may be wrong. Perhaps it was a precursor to Colossus that killed her.

[Anomalocaris]

Self driving cars are actually very good. They have been around for a while now, this is the first case of a death.

What I really want to know is why the human driver didn’t brake. The whole reason there’s still a human behind the wheel is to act when the car fails. There’s clearly still work to do, but we also shouldn’t overreact to one case.

It is often found that during pilot training, inexperienced flight instructors tend to delay taking over the control of the aircraft for safety when the student pilot does dangerous things. The instructors are trained to be vigilant for such action and promptly take control of the aircraft when the student pilot screw up. But when someone or something is apparently operating a vehicle normally, it seems to be reflexive for humans to give the controller more latitude then is warranted when situation suddenly change.

[vorlon13]

I think the news report this evening said the car was going around 40. There might not have been much reaction time available. If the accident was inevitable with a human or computer operator that would be significant. Additionally, I assume an Uber car in automatic driving mode would record sensor inputs and control settings and such. Perhaps a video of what the (human) driver would have seen.

[AFTT47]

Self driving cars are actually very good. They have been around for a while now, this is the first case of a death.

What I really want to know is why the human driver didn’t brake. The whole reason there’s still a human behind the wheel is to act when the car fails. There’s clearly still work to do, but we also shouldn’t overreact to one case.

It is often found that during pilot training, inexperienced flight instructors tend to delay taking over the control of the aircraft for safety when the student pilot does dangerous things.   The instructors are trained to be vigilant for such action and promptly take control of the aircraft when the student pilot screw up.   But when someone or something is apparently operating a vehicle normally, it seems to be reflexive for humans to give the controller more latitude then is warranted when situation suddenly change.

Yes, of course. Complacency is natural for a human - part of the attraction of self-driving cars in the first place. When the computer gets it right 99.9% of the time, the human overseer loses interest and is not a reliable fail-safe.

I've always thought that the most difficult part of developing this tech is going to be identifying that infinitesimally small time when they will fail so that you can make corrections to the program. You can't really be aware of the flaw until it actually occurs. You hope your human overseer will kick in and prevent the actual accident but they are going to get complacent because they are human.

[Minimalist]

Somehow, I doubt that Addison said that.