Habryka [Deactivated] answers What predictions from theoretical AI Safety research have been confirmed by empirical work?