Translated data: Google's team has introduced a new universal visual encoder called VideoPrism, which, based on extensive pre-training with massive video data and text pairs, has set 30 new state-of-the-art records. This model is capable of handling various video understanding tasks, including classification, localization, retrieval, captioning, and question answering. Google's VideoPrism demonstrates strong versatility and generalization capabilities, bringing significant breakthroughs to the field of video technology.