Translation: The research teams from MIT and Stanford have introduced a novel machine learning framework that enables large AI models to actively ask questions to humans, thereby better understanding user preferences. This method, known as GATE (Generative Active Task Elicitation), demonstrates superior effectiveness compared to traditional prompt engineering and supervised learning. This research has sparked discussions on prompt engineering, with some arguing that it will not disappear but will undergo a transformation in form. This technology may find expansion in fields such as healthcare and law, which require complex decision-making processes.