Core Insights - Beijing is set to have over 2,400 artificial intelligence companies by 2024, with a core industry scale nearing 350 billion yuan, accounting for half of the national total, solidifying its status as the "AI Capital" of China [1][2] Group 1: Innovation and Development - Artificial intelligence is recognized as a strategic technology driving a new wave of technological revolution and industrial transformation, significantly altering human production and lifestyle [1] - Beijing has established a systematic layout in AI technology innovation, application demonstration, and innovation ecology, contributing to its leading position in the AI sector [1] - The city boasts 21 national key laboratories and over 40% of the nation's top talent, focusing on areas such as large models and AI safety [1][2] Group 2: Policy and Infrastructure - Beijing has introduced various policies, including a three-year action plan for embodied intelligence and initiatives to integrate AI with new materials, accelerating the development of new AI sectors [2] - The city is enhancing its original innovation capabilities, producing groundbreaking technologies like the world's first optical training chip and establishing a national first 64-card super-node computing server [2] - A comprehensive plan for computing infrastructure is in place, with an expected addition of 8,620 PetaFLOPS of computing power by 2024, bringing the total to over 33,000 PetaFLOPS [2] Group 3: Ecosystem and Collaboration - An open innovation ecosystem is rapidly forming, with international AI seminars and conferences attracting global participation, indicating a strong collaborative spirit [3] - Future plans include a focus on original innovation and the development of disruptive technologies in fields like optical computing and brain-like intelligence [3][4] - Beijing aims to create a first-class development environment by enhancing talent cultivation and improving financial services for AI enterprises [4] Group 4: Global Engagement - The city is committed to deepening global cooperation, positioning itself as a "global open-source capital" and enhancing its international influence in AI [4] - Efforts are underway to implement global AI governance initiatives and foster international dialogue on AI safety and research collaboration [4]

经过几年的快速发展，面壁智能已形了成全模态、多模态与基座模型组成的完整矩阵，持续刷新端侧大 模型能力边界。MiniCPM凭借以小博大、高效低成本的特性享誉全球，接连实现端侧ChatGPT、GPT- 4V、GPT-4o时刻，并获评2024年Hugging Face"最多下载、最受欢迎中国大模型"。截至目前，面壁小钢 炮MiniCPM系列平台下载量累计破1000万。 在应用侧，面壁智能发布的"小钢炮超级助手cpmGO"，是全球首个落地车端的纯端侧智能助手；面壁 智能作为联合研发团队参与的"法信法律基座大模型"也已在最高人民法院发布；与清华大学联合推出 AI 学生成长助手"清小搭"，为清华大学所有本科生配备具有智能问答、智能伴学功能的专属智能助 手…… 据介绍，本轮融资的完成，将进一步为面壁智能构筑高效大模型技术、产品壁垒、加速行业赋能与生态 拓展奠定坚实基础，协同产业上下游推动「端侧大脑」在千行百业规模化应用。 近日，面壁智能完成新一轮数亿元融资，本轮融资由洪泰基金、国中资本、清控金信和茅台基金联合投 资，自 2024年以来，面壁智能已连续完成3轮融资。 2024年9月，面壁智能发布小钢炮 MiniCPM 3.0 ...

5 月 7 日，由 GOSIM、CSDN 和 1ms.ai 联合主办的全球开源技术盛会——GOSIM AI Paris 2025 在法国巴黎迎来了大会第二日的精彩议程。 延续首日的热烈氛围 ， 来自全球的 AI 专家、开发者和 产业代表齐聚一堂，围绕 AI 技术的最新趋势与实践展开深入探讨。 从底层技术架构到前沿应用场景，本次大会围绕 AI 模型、AI 基础设施、AI 应用及具身智能等核心 主题设立技术论坛，全方位覆盖 AI 领域的关键议题。此外，Spotlight Talks、PyTorch Day 等特色 活动也为技术开发者搭建了一个思想碰撞与成果展示的技术舞台。 多模态统一架构。尽管"统一架构"仍有不少技术争议，但多个顶级实验室的研究已显示出明显趋 势，即试图将多模态建模能力整合进统一框架。他特别提到 Meta 于去年年底发布的 BLT（Byte Latent Transformer）架构，展示了这一方向的可行性。Zhipeng Huang 认为，2025 年有望 诞生具突破性的统一模型，对现有主流架构发起挑战。 高效注意力机制演进。线性注意力、动态稀疏注意力和混合注意力等机制正快速发展。 二阶优化走向 ...