图片来源： 阶跃星辰 段楠近日在其 GitHub 主页更新履历：现已出任 京东探索研究院视觉与多模态实验室负责人 ，带领团队专注于视觉与多模态基础模型的研发工作。 加入京东前，段楠曾担任 阶跃星辰 Technical Fellow、视频模型负责人 ，也曾在微软亚洲研究院担任 自然语言计算团队资深首席研究员 ，拥有深厚的跨模 态与AI模型研究背景。 段楠近日从阶跃星辰离职，业内普遍解读为阶跃星辰在视频模型方向投入缩减的信号之一。作为公司多模态战略的关键人物之一，段楠的离开，或预示着 这家明星初创正在悄然调整技术重心。 在"六小龙"中，阶跃星辰曾是多模态路线最鲜明的代表：成立两年内推出了22款自研基础模型，其中多达16款为多 模态模型，覆盖文本、语音、图像、视频、音乐与推理等方向，占比超过七成，重仓姿态可见一斑。 不仅如此，阶跃星辰曾一度被视为"六小虎"中最为稳固的存在—— 是唯一一家未曾曝出核心高管离职消息的企业 ，也因此被外界认为团队凝聚力强、 战略方向清晰。而 段楠的出走标志着这一记录被打破。 | 所在公司 | 人名 | 职位 | 离职时间 | 去向 | 介绍 | | --- | --- | --- | - ...

Group 1 - The core viewpoint emphasizes that AI should not replace humans but rather understand and serve them better, becoming a daily necessity for the public [1][6] - Ma Yun and Xu Li both highlighted the importance of AI integrating into everyday life, making it accessible to the general public [1][6] - The Stanford University AI Index Report indicates that the performance gap between top AI models in China and the US has significantly narrowed from 17.5% in 2023 to 0.3% in 2024 [10] Group 2 - Xu Li announced the upgraded "SenseNova V6" model, which features over 600 billion parameters and surpasses GPT-4 in data analysis capabilities [7] - The new model supports understanding long videos of up to 10 minutes, marking a significant advancement in multi-modal AI capabilities [7] - The report also notes that in 2024, the US had 40 notable models while China had 15, indicating a competitive landscape in AI development [10] Group 3 - The application of AI in insurance claims processing is highlighted, showcasing its ability to analyze and organize information effectively, addressing issues of information asymmetry [5] - The integration of multi-modal capabilities is seen as essential for achieving general artificial intelligence, as it encompasses text, images, and videos in a unified processing system [3][5] - The push for AI education in Guangdong province reflects a broader trend of digital transformation in education, with initiatives aimed at enhancing AI literacy among students and teachers [8]