Uh oh, another Tesla autopilot crash fatality

[MetroMPG]

Supposedly the car checks frequently for driver input via steering. If there's no input and the driver ignores the warnings and alarms, the car brings itself to a stop.

[sendler]

Tesla would be wise to disable all autopilot requests and continue testing only with paid employees.

[jamesqf]

Originally Posted by RedDevil

No driver can shed responsibility to any control system in any kind of car.

Unfortunately, they can and do - as they did in both these instances - shed their actual responsibility. As I said, I am not a lawyer, so I don't know how the legal responsibility issue might play out in the courtroom, but historical evidence (McDonald's hot coffee, for instance?) is not encouraging for Tesla.

Moreover, shedding driver control & responsibility is exactly what Musk and other developers of self-driving car systems have been hyping all along. Would a jury fault a Tesla buyer for believing the hype rather than what's in the manual?

[jamesqf]

Originally Posted by sendler

Tesla would be wise to disable all autopilot requests and continue testing only with paid employees.

They'd be even wiser to stop trying to develop it altogether, and devote their limited resources to building cars, batteries, and charging networks. If Google or someone ever gets the technology working, they can license it, just like all the other car makers.

[sendler]

Originally Posted by jamesqf

Moreover, shedding driver control & responsibility is exactly what Musk and other developers of self-driving car systems have been hyping all along.

It's not just the auto makers hyping self driving cars. Whole government agencies all over the world are pushing for this. They think a car can do a much safer job of driving than a person. Unbelievable.

[redpoint5]

There have been several big commercial airline crashes that have been caused by a pilot accidentally pressing on the flight stick, or bumping into some other button, and autopilot disengaging while giving the alert tone and visual indication that the plane was in manual flight mode. The pilot and copilot still failed to comprehend this until the planes attitude is off. The pilot then gets disoriented and fails to regain attitude control and crashes. These disasters could have been entirely avoided if the auto pilot was re-engaged.

If the facts in this Tesla crash are true, then liability rests entirely on the human driver. He took manual control when he provided steering input and throttle control. Can you imagine the backlash and liability Tesla would get if a driver took manual control of the wheel, but was overridden by the autopilot and then a subsequent crash resulted?

Autopilot always has to surrender control to the human pilot when they give an input to the controls. Nobody would ride in a car that can override our driving inputs.

[freebeard]

Google looked at self-driving cars, and their research scared them so much they took out the steering wheel and brake completely.

So it looks like a future mix of completely autonomous and manual cars. I'm Okay with neighborhood shuttle buses like Olli and interstate trucks and buses, but mixed urban traffic will be... law abiding?

[jamesqf]

Originally Posted by freebeard

...but mixed urban traffic will be... law abiding?

Frankly, I don't care much about urban driving. I think the big problems will start to show up when those urbanite-designed and tested self-driving cars start trying to travel outside the confines of their urban sandboxes.

We've already seen instances of human drivers who've blindly followed their GPS into deserts, over cliffs, into rivers, and across bridges that obviously wouldn't carry the commercial trucks they were driving. That's what happens with just a small fraction (at least I hope it's small :-)) of the population dumb enough to blindly follow a machine's instructions. What happens when the follower is another machine (or part of the same one), and the blind obedience is built in?

[freebeard]

Peter Diamandis

And critically, all of the data processed by these sensors is fed back to a central hub so that all Tesla cars can learn from it, called Fleet learning.

The data is aggregated (anonymously) into maps that let the central Autopilot system see the precise paths that cars take, and don’t take.

“Each driver is effectively an expert trainer in how Autopilot should work,” Elon explained, and the network of Tesla vehicles is constantly learning more about where cars do and don’t actually travel.

There's a reason it's called the 'bleeding edge'.

[MetroMPG]

Originally Posted by jamesqf

Reminds me of the old airplane joke. Airline pilot message to maintenance: "Autoland system on this plane is very rough". Maintenance reply: "Plane does not have autoland".

HA! I'll be re-telling that one.