[mieseratte]

As an occasional commuter cyclist, I'm well aware of the position cars and drivers hold in our social pecking order. It's a well known "joke" that if you want to get away with murder, you do it with a car.

While we should not absolve Uber, including the driver who was too busy looking at his phone to intervene, I do wonder if a standard driver would have also been deemed not at fault by the police. The answer is likely, "yes".

[simion314]

> I do wonder if a standard driver would have also been deemed not at fault by the police. The answer is likely, "yes".

But it seems, we do not for sure, that the sensors completely failed and on top of that the human driver also failed, so the question is :are this tests safe? next time the failure could happen on a crosswalk , IMO the question of who is to blame is not as important then the question if is safe to do this tests on public roads with such poor hardware and software.

[avs733]

>While we should not absolve Uber, including the driver who was too busy looking at his phone to intervene, I do wonder if a standard driver would have also been deemed not at fault by the police. The answer is likely, "yes".

Add in the characterization of the victim that has been going on. If the car had hit an ASU student it would have been assumed they were drunk/on the phone. If it had been a Mormon missionary (not uncommon in AZ) there would be a lot more focus on the car/driver. Instead, they hit a homeless/low value person. The discourse reflects that.

[masklinn]

> including the driver who was too busy looking at his phone to intervene

I've yet to get any information on one thing though, are safety drivers operating under the assumption that the car works at SAE2 or at SAE3? Because if it's the latter, the driver has no cause to keep looking at their phone. If it's the former, the car should have a deadman's switch to ensure the driver stay alert.

An other thing that is not clear is whether they were looking at their phone or at instrumentation (e.g. telemetry or the like).

[ghaff]

The safety drivers should theoretically be operating under the assumption that the self-driving system can fail randomly at any moment. Of course, humans are not wired as reliable backups to handle random split-second problems when the system, in fact, works correctly most of the time.

[masklinn]

> The safety drivers should theoretically be operating under the assumption that the self-driving system can fail randomly at any moment.

Then the car should be equipped with a dead man's switch / vigilance devices to ensure the driver pays attention, trains have been equipped with these equipments for decades.

> Of course, humans are not wired as reliable backups to handle random split-second problems when the system, in fact, works correctly most of the time.

Indeed, but again that is a long-known issue and we've had solutions for a long time.