3 ms·
In a fatal crash, Uber’s autonomous car detected a person, but chose to not stop
- consumer451 9y agoMany years ago I was coding the UI for one of the first touchscreen check-in systems at an airline. The guy in the next cube was doing weight and balance software for the actual flights. I realized that I could not handle the pressure of thinking that a bug in my code could lead to a crash, no matter how remote the chances actually were. I'm sure that some of you work on code where lives are at stake, how do you deal with the possibility of truly fatal bugs?
- natch 9y agoA while back some academics at MIT designed some surveys for probing what tradeoffs humans would make if they had a chance to decide between bad outcomes in a vehicle fatality incident. https://www.technologyreview.com/s/542626/why-self-driving-cars-must-be-programmed-to-kill/ https://www.technologyreview.com/s/542626/why-self-driving-c... This shows that the menu of ethical dilemmas posed at that time was not complete. In addition to "should I veer left, avoiding the grandmother and grandfather couple, but killing the pregnant mother" the survey questions should have also included some questions like "should I brake hard to avoid killing the homeless person detected with 40% probability, or should I assume the detection is a false positive, in order to maintain a smooth ride of luxurious comfort for the occupants?"
- privong 9y agoLots of discussion here already: https://news.ycombinator.com/item?id=17014807 https://news.ycombinator.com/item?id=17014807
- consumer451 9y agoSorry, I looked but didn’t see it. Thanks.