The article focuses on trains, but the same thing could be said for autonomous vehicles. When one crashes it is national news because they are so rare. People who say autonomous vehicles won't be viable until the crash rate is zero are setting the bar far too high. The threshold should be fewer accidents and/or fewer fatalities per million miles driven than human drivers. In theory system improvements on the autonomous cars could further reduce those figures, while reducing the accidents involving human factors is very difficult. You can't change people that much. Most of the foreseeable improvements are basically the same ones you would need for autonomous vehicles anyway, like better sensors around the car to detect obstacles.