The 'Trolley Problem' — 40m answers

[Wetstuff]

Since there's only me at the moment ...and 112 watchers, maybe they'll be interested in this?

https://en.wikipedia.org/wiki/Trolley_problem





Jim

[PonoBill]

Very interesting. One of the reasons Tesla has backed away from their timeline prediction for full autonomy is a subset of this problem--explaining to regulators how these kinds of choices would be made. The concern creates a stalking horse for the regulators that is very hard for the proponents to deal with. It doesn't matter how many lives autonomous cars in the aggregate will save (a lot) if one runs over your kid to save your wife. Either choice can be painted as "wrong".

On another note, Tesla announced a profit today. The Model 3 is the largest selling car in America by revenue and the second largest selling in volume. Interesting.

[exiled]

I'm struck by the phrase "Hard cases make bad law." It seems like the most palatable solution would be a routine that is simple and transparent. Something like: If an obstruction is detected, brake and swerve into clear path. If no clear path is available, brake and hope for the best. You wouldn't expect more than that from a human driver anyway and as a pedestrian you can predict the car's behavior. Plus if your car can start detecting people and anticipating impact, you are not far from employing new safety tech like a bumper deployed airbag to pad the pedestrian impact.

[yugi]

[eastbound]

not sure how the wall st clip relates

douglas happens to be correct about the frequesnt uselessness of over-pedigreed prima donnas

[PonoBill]

People have a natural concern about machines running around unattended on public roads that can kill them and/or their families. Even people who haven't thought deeply about autonomous vehicles voice that concern. A person driving a car, like those two ditzy distracted blondes, are not only unaware of the potential accident until it's too late to avoid, but also can't control their vehicle well enough to take deliberate action. Any autonomous car can. The car might make mistakes in recognition (though that potential is rapidly declining as data is added) but it will always see the unfolding problem and has, in relative terms, all the time in the world to react properly since it's decisions happen at machine speed. It can take decide among literally myriads of responses and choose one, limited only by the physics of the car's momentum and performance envelope.

Then the question is, which decision is right?

A human driver, seeing a dog run out into the road slams on the brakes. The autonomous car knows that there is a ten-ton box truck close behind it and what speed they are traveling at, can guess with reasonable accuracy how quickly it can stop and how likely it is to stop, assess all alternative paths and decide it should run the dog down.

As the passenger, how do you judge that action?

Now bump the stakes higher. The car is full of passengers, one kid runs into the street and there are only two choices. Brake hard and accept an inevitable rear-end collision or run over the kid.

If the car is being driven by the ditzes, the accident that results in the kid being run over is truly an accident. They don't have the attention and time to decide, the reflexes to avoid or the skill to pull it off. The autonomous car does and it is no accident that it runs over the kid and spares the car's passengers from potential death or injury.

Even if autonomous cars eventually reduce the carnage by a huge margin, they have to become common enough to numerically matter. Cold-blooded least harm decisions are going to make a HUGE fuss, and the decisions have to be connected to the morality of the host country.

Humans have a problem with change--they expect all the benefits and none of the problems from any change. People constantly say electric cars are not environmentally sound because they are fueled with electricity from coal plants, completely or more probably, willfully, ignoring the fact that they do not have to be. Or the fact that  4.5 gigawatts of coal, 2 percent of 2016 U.S. coal capacity, are being retired and only one new coal plant is planned for the next decade--and it is unlikely to be built. Instead, we will add 11 GW of natural gas and 8.5 GW of wind. So hey, that argument is bullshit squared.

I squirmed watching bunhead boy lecture on the methodology of the research, but it's good stuff, and necessary. People are weird.