豆包可以跟你打视频了,陪我看《甄嬛传》还挺懂!难倒一众AI的“看时钟”也没难倒它
量子位·2025-05-26 08:18

Core Viewpoint - The article discusses the advancements in domestic AI technology, particularly focusing on the new video call feature of the "Doubao" AI, which can accurately tell the time and engage in real-time conversations while watching videos [1][3][4]. Group 1: AI Capabilities - The domestic AI can accurately report the time during a video call, demonstrating significant improvement over previous models that struggled with such tasks [2][3]. - The AI integrates internet search capabilities, enhancing the accuracy and timeliness of its responses to current events and trending topics [6][7]. - The new feature includes subtitles, allowing users to view the conversation history, which adds to the interactive experience [9]. Group 2: Practical Applications - The AI can serve as a companion for watching shows, accurately identifying scenes and providing commentary, as demonstrated with the show "Zhen Huan Zhuan" [16][18]. - It can assist in cooking by recognizing ingredients and providing detailed cooking instructions, showcasing its practical utility in everyday tasks [20][22]. - The AI is capable of solving academic problems, such as physics questions, and can assist with understanding complex topics like calculus, highlighting its educational applications [23][34]. Group 3: Underlying Technology - The "Doubao Visual Understanding Model" powers the AI's capabilities, featuring strong content recognition abilities that allow it to identify various elements in images [24][25]. - The model excels in understanding and reasoning, enabling it to perform complex logical calculations and provide clear problem-solving strategies [33][34]. - The AI's detailed visual description and creative capabilities contribute to its effectiveness in real-time interactions, making the user experience engaging and informative [35][36].