单词学习：![[0a0dbf74-acf8-44b0-b503-4c9a87dfa2b6.mp3]]

新闻播报：![[b752f813-1cb5-468a-b436-4e07c5fdde7e.mp3]]

DeepMind CEO：中国AI仅落后数月，但缺乏颠覆性创新

简报：

• Google DeepMind CEO哈萨比斯（Demis Hassabis）对中国AI发展现状发表了看法。
• 他肯定了中国AI的技术追赶速度，认为与世界前沿水平的差距可能仅有数月。
• 但哈萨比斯同时强调，在实现“颠覆性创新”（disruptive innovation）上，中国与领先者之间仍存在关键差距。
• 近期，中国AI产业持续发展，企业数量已超6200家，美团等公司也发布了具有SOTA（顶尖水平）能力的开源大模型。

相关链接：

• DeepMind CEO哈萨比斯罕见肯定中国AI进展：仅落后数月，但“颠覆性创新”仍是关键差距
• OpenAI 发布 GPT-5.2-Codex 编程模型，API 正式开放
• 全球首个开源“重思考”模型：美团 LongCat-Flash-Thinking-2601 发布，工具调用能力登顶开源 SOTA
• 我国人工智能企业数量已超 6200 家，AI 大模型融入千行百业

DeepMind CEO: China's AI Only Months Behind, But Lacks Disruptive Innovation

Briefing:

• Demis Hassabis, CEO of Google DeepMind, shared his views on the current state of AI development in China.
• He acknowledged China's rapid pace in catching up technologically in AI, suggesting the gap with global frontier levels might only be a few months. 🚀
• However, Hassabis also emphasized that a critical gap still exists between China and leading innovators in achieving "disruptive innovation." 🤔
• Recently, China's AI industry has seen continuous growth, with the number of AI enterprises exceeding 6,200. Companies like Meituan have also released open-source large models with SOTA (state-of-the-art) capabilities. 🤖

Related Links:

• DeepMind CEO Demis Hassabis Unusually Praises China's AI Progress: Only Months Behind, But "Disruptive Innovation" Remains the Key Gap
• OpenAI Releases GPT-5.2-Codex Programming Model, API Officially Open
• World's First Open-Source "Re-thinking" Model: Meituan LongCat-Flash-Thinking-2601 Released, Tool Calling Capability Tops Open-Source SOTA
• China's Number of AI Enterprises Exceeds 6,200, AI Large Models Integrated Across Industries

我国AI企业数量突破6200家，大模型正赋能各行各业

简报：

• 据央视新闻报道，截至目前，我国人工智能企业数量已超6200家，AI大模型正融入千行百业并拓展应用场景。
• 据中国电子信息产业发展研究院数据，2025年我国人工智能核心产业规模已突破1万亿元。
• AI大模型已应用到金融、医疗、制造等多个行业，将传统经验决策升级为“数据+模型”驱动的智能决策。
• 此前数据显示，截至2025年9月，我国人工智能企业数量为5300余家，占全球总数的15%。

相关链接：

• 我国人工智能企业数量已超 6200 家，AI 大模型融入千行百业
• 全球首个开源“重思考”模型：美团 LongCat-Flash-Thinking-2601 发布，工具调用能力登顶开源 SOTA

China's AI Enterprise Count Surpasses 6,200, Large Models Empowering Diverse Sectors

Briefing:

• According to CCTV News reports, as of now, the number of AI enterprises in China has exceeded 6,200, with AI large models integrating into countless industries and expanding application scenarios. 🚀
• According to data from the China Center for Information Industry Development (CCID), the scale of China's core AI industry has already surpassed 1 trillion yuan by 2025.
• AI large models have been applied across various industries such as finance, healthcare, and manufacturing, upgrading traditional experience-based decision-making to "data + model"-driven intelligent decision-making. 💡
• Previous data indicated that as of September 2025, China had over 5,300 AI enterprises, accounting for 15% of the global total. 📈

Related Links:

• 我国人工智能企业数量已超 6200 家，AI 大模型融入千行百业
• 全球首个开源“重思考”模型：美团 LongCat-Flash-Thinking-2601 发布，工具调用能力登顶开源 SOTA

微软发布Copilot Studio扩展，AI智能体开发迈向专业化

简报：

• 微软正式推出 Visual Studio Code 版 Copilot Studio 扩展，将 AI 智能体开发从低代码网页端迁移至专业集成开发环境（IDE）。
• 该扩展实现了与 Git 版本控制系统和 CI/CD 流水线的深度集成，支持开发者遵循标准软件工程工作流进行协作、测试与部署。
• 新工具提供结构化定义、语法高亮和代码补全等功能，并能与 GitHub Copilot 等AI编程助手协同，以提升开发效率和降低错误率。

相关链接：

• 微软 Copilot Studio 扩展正式上线，深度集成 Git 与 CI/CD 流水线

Microsoft Launches Copilot Studio Extension, Professionalizing AI Agent Development 🚀

Briefing:

• Microsoft has officially released the Copilot Studio extension for Visual Studio Code, shifting AI agent development from low-code web interfaces to a professional Integrated Development Environment (IDE). 💡
• The extension offers deep integration with Git version control systems and CI/CD pipelines, enabling developers to follow standard software engineering workflows for collaboration, testing, and deployment.
• New tools provide features like structured definitions, syntax highlighting, and code completion. They also synergize with AI coding assistants such as GitHub Copilot to boost development efficiency and reduce error rates. 🛠️

Related Links:

• Microsoft Copilot Studio Extension Officially Live, Deeply Integrated with Git and CI/CD Pipelines

人形机器人高空作业新突破：开普勒实现远程遥控焊接

简报：

• 开普勒于1月15日宣布，其人形机器人完成了全球首例“人机协作”高空焊接作业。
• 作业过程中，地面操作员佩戴VR头显，通过沉浸式全身遥操系统，1:1实时控制20米高空的开普勒K2大黄蜂人形机器人。
• 该机器人以毫米级精度复刻操作员动作，可持续8小时不间断作业，并通过力反馈系统将阻力实时传递给操作员。
• 系统支持最长2000公里的超远距离遥控，并能通过反复遥操训练让机器人自主优化动作策略。

相关链接：

• 开普勒人形机器人实现全球首例“人机协作”高空焊接作业
• 开普勒人形机器人实现全球首例“人机协作”高空焊接作业
• 开普勒人形机器人实现全球首例“人机协作”高空焊接作业

Humanoid Robot Breakthrough in High-Altitude Operations: Kepler Achieves Remote Controlled Welding 🚀

Briefing:

• Kepler announced on January 15th that its humanoid robot successfully completed the world's first "human-robot collaborative" high-altitude welding operation. 🤖
• During the operation, a ground operator, wearing a VR headset, controlled Kepler's K2 'Bumblebee' humanoid robot 20 meters above ground in 1:1 real-time via an immersive full-body teleoperation system.
• The robot replicated the operator's movements with millimeter-level precision, could operate continuously for 8 hours, and transmitted resistance back to the operator in real-time through a force feedback system.
• The system supports ultra-long-distance remote control up to 2000 kilometers and allows the robot to autonomously optimize its action strategies through repeated teleoperation training. ✨

Related Links:

• Kepler Humanoid Robot Achieves World's First "Human-Robot Collaborative" High-Altitude Welding Operation
• Kepler Humanoid Robot Achieves World's First "Human-Robot Collaborative" High-Altitude Welding Operation
• Kepler Humanoid Robot Achieves World's First "Human-Robot Collaborative" High-Altitude Welding Operation

智谱华为联手，国产芯训练图像模型登顶Hugging Face

简报：

• 智谱联合华为开源了新一代图像生成模型 GLM-Image，该模型是首个完全依托国产芯片和AI框架完成全程训练的SOTA多模态模型。
• 模型基于华为昇腾 Atlas 800T A2 设备和昇思 MindSpore AI 框架进行训练，验证了在国产算力上训练出顶尖模型的可行性。
• GLM-Image 开源不到24小时，即登顶全球AI开源社区 Hugging Face Trending 榜单第一名。
• 该模型尤其擅长汉字生成任务，在复杂视觉文本生成（CVTG-2K）和长文本渲染（LongText-Bench）榜单上获得开源模型第一。

相关链接：

• GitHub 开源地址
• Hugging Face 开源地址

Zhipu AI and Huawei Join Forces: Domestic Chip-Trained Image Model Tops Hugging Face

Briefing:

• Zhipu AI, in collaboration with Huawei, has open-sourced GLM-Image, a new generation of image generation model. This model is the first SOTA (State-of-the-Art) multimodal model to be entirely trained using domestic chips and AI frameworks. 🚀
• The model was trained on Huawei Ascend Atlas 800T A2 devices and the MindSpore AI framework, proving the feasibility of training top-tier models on domestic computing power.
• Less than 24 hours after its open-source release, GLM-Image soared to the top spot on the Hugging Face Trending list, the world's leading AI open-source community. 🥇
• The model particularly excels in Chinese character generation tasks, securing the #1 position among open-source models on both the Complex Visual Text Generation (CVTG-2K) and LongText-Bench leaderboards. ✨

Related Links:

• GitHub Open Source Link
• Hugging Face Open Source Link

B站推出AI营销工具，瞄准年轻消费者

简报：

• B站（Bilibili）发布了一款一站式AI营销工具，旨在通过人工智能技术，帮助品牌和广告主更好地发力并触达年轻人的消费市场。

相关链接：

• 相关报道3

Bilibili Launches AI Marketing Tool, Targeting Young Consumers 🎯

Brief:

• Bilibili (B站) has launched a one-stop AI marketing tool, designed to empower brands and advertisers to better strategize and effectively reach the young consumer market through artificial intelligence technology. 🚀✨

Related Links:

• Related Report 3