Core Insights - The 2025 World Artificial Intelligence Conference (WAIC 2025) held in Shanghai focused on the theme "Intelligent Era, Shared Future," bringing together global experts and representatives to discuss new opportunities in AI development [2] - Yan Junjie, CEO of MiniMax, emphasized that AI companies should not be viewed merely as extensions of internet companies but as infrastructure enterprises focused on capability building, providing enhanced productivity for society [2] Company Developments - MiniMax, founded in 2022, has developed several multimodal general models, including MiniMax M1, Hailuo-02, Speech-02, and Music-01, capable of processing text, audio, images, video, and music [7] - The company plans to launch its first full-stack general intelligence product, "MiniMax Agent," during the conference, designed to handle long-term complex tasks with capabilities in task planning, sub-task breakdown, and multi-thread execution [6][7] - MiniMax's products have served approximately 157 million individual users and over 50,000 enterprises and developers across more than 200 countries and regions [7] Industry Trends - AI is transitioning from task assistance to task leadership, increasingly penetrating complex environments, showcasing its ability to understand systems, collaborate on multiple tasks, and learn with goal orientation [3] - The efficiency of AI models has improved significantly, with training costs stabilizing, indicating that future R&D advantages will rely more on effective experimental design and team capabilities rather than solely on computational power [5] - The AI landscape is evolving towards a decentralized, multi-center development model, where various organizations can align models according to their preferences, leading to diverse systems that emphasize different aspects such as code execution, emotional interaction, and creativity [5][6] - The trend towards "inclusive AI" is evident, as both training and inference costs are decreasing, allowing for broader deployment of AI across various social scenarios [6]

Core Insights - The article discusses the launch of new AI models by domestic large model manufacturers, specifically highlighting MiniMax-M1 and Kimi-Dev-72B as significant advancements in the field of open-source AI models [1][9]. Group 1: MiniMax-M1 - MiniMax-M1 is introduced as a long-context reasoning LLM capable of handling an input of 1 million tokens and an output of 80,000 tokens, making it one of the most powerful models in terms of context length [2][19]. - The model demonstrates exceptional capabilities in interactive applications, such as creating web applications and visualizing algorithms, with a focus on user-friendly UI components [5][8]. - MiniMax-M1 has been trained using a novel reinforcement learning algorithm called CISPO, which optimizes model performance by focusing on important sampling weights rather than token updates, achieving faster convergence compared to previous methods [20][23]. - The model's performance in various benchmarks shows it surpasses other open-weight models, particularly in software engineering and long-context tasks, with a notable score of 56.0% on the SWE-bench Verified benchmark [29][25]. Group 2: Kimi-Dev-72B - Kimi-Dev-72B is presented as a powerful open-source programming model that achieved a new state-of-the-art (SOTA) score of 60.4% on the SWE-bench Verified benchmark, showcasing its capabilities in code generation [10][37]. - The model employs a collaborative mechanism between BugFixer and TestWriter roles, enhancing its ability to fix bugs and write tests effectively [40][45]. - Kimi-Dev-72B underwent extensive mid-training using high-quality real-world data, which significantly improved its performance in practical error correction and unit testing [41][42]. - The model's design includes a unique outcome-based reward mechanism during reinforcement learning, ensuring that only effective code fixes are rewarded, thus aligning with real-world development standards [43][44].

京沪两地网信部门近日公布整治AI技术滥用的阶段性进展。这些进展涵盖清理违规AI产品、强化AI内 容标识等方面，基本上对标了中央网信办4月底专项行动第一阶段的任务清单。 落地情况来看，官方信息显示，北京市目前已有近60家企业已落实标识要求；微博、抖音等平台实现生 成内容"显著标识+元数据水印"双标注。 在训练语料治理上，百度对存量数据进行整体清洗，使用211个权威网站的正能量数据源；百川智能停 止使用存在疑问的数据来源，制定严格的网络爬取规范，确保数据合规合法。 上海属地企业方面，MiniMax、阶跃星辰、通义、小红书、哔哩哔哩、Soul等已基本完成显式标识规范 上线，加快开发上线隐式标识和传播链互认验证；小红书牵头编制图片模态元数据标识实践指南。此 外，上海网信部门还在5月中旬指导成立AIGC标识生态联盟，推进标识互认、能力共享与标准协同。小 红书、MiniMax等均是该联盟的成员。 而在医疗、金融、未成年人等重点领域的风险防控上，针对"AI荐股"，智谱明确禁止生成具体股票推 荐、基金购买建议，避免诱导高风险投资；快手对医疗回答明确标注"不替代专业医疗建议"。未成年人 保护方面，猿辅导背后的猿力科技对未成年人 ...

Core Insights - The "AI Super Factory" known as "Mosu Space" is rapidly reshaping the AI industry landscape in Shanghai, with nearly 400 large model enterprises gathered in a 1.5 square kilometer area, accounting for 61% of the city's total registered large models [2][3][9] - Mosu Space aims to become a global AI incubator, providing a comprehensive ecosystem that includes computing power, data, and policy support to foster innovation and collaboration among AI companies [6][7][9] Industry Overview - The area has attracted leading companies such as Jieyue Xingchen, Wuwen Xinqun, and SenseTime, forming a complete industrial chain from computing chips to applications in various sectors like healthcare and finance [2][4] - The ecosystem supports a wide range of applications, with companies like MiniMax and Jieyue Xingchen releasing advanced models and applications that have gained global traction [4][5] Innovation and Collaboration - Mosu Space promotes a collaborative environment where companies can easily access resources and partners, enhancing their innovation capabilities [6][7] - The space has established five functional platforms to support startups, including computing power scheduling and financial services, ensuring a "one-stop" service for incoming enterprises [6][9] Future Aspirations - The Shanghai Xuhui district has ambitious plans to expand its AI ecosystem, aiming for significant growth in the number of large model enterprises and innovation spaces [9][10] - Mosu Space is positioned to become a key player in the global AI landscape, with initiatives to enhance its international presence and attract global AI companies [9][10]

以下文章来源于IT时报 ，作者孙妍 IT时报 . 做报纸，也懂互联网，这里是《IT时报》(IT Times)微信版。作为上海一份IT类周报的新媒体产品，这 里汇聚了关注全球IT业的魔都资深IT记者。我们追求原创独家新锐，以及读视听多种表达方式。ps. 使 用IT产品有问题？留言与编辑互动。 人工智能作为国家战略级重点发展领域，正在以惊人的速度自我迭代，更推动了千行百业转型升级。 成立于2014年的商汤，于2017年落户上海市徐汇西岸，见证并参与了上海人工智能的发展。在它落 沪这一年，上海市人民政府办公厅印发《关于本市推动新一代人工智能发展的实施意见》通知。 从AI视觉识别时代一直走到大模型时代，前瞻性布局让商汤走在了正确的道路上。 2025年是大模型应用元年。商汤徐立与阿里马云隔空对话时，不约而同地提到，AI要成为老百姓 的"日用品"。毕竟，性价比才是大模型应用生死线。今年4月10日，商汤宣布发放1亿元算力代金 券，加速大模型商业化落地。 除了算力这一基础设施外，商汤更担当着基础大模型的角色，不断突破性能，要与世界同行比肩。4 月10日，商汤发布日日新SenseNova V6大模型体系，多模态推理能力对标Op ...