![[c05a3057-ab3b-44eb-a9be-2749adcfbebf.mp3]]

希捷提醒AI存储扩张或加剧行业碳排放隐忧

简报：

• 随着AI应用带来海量数据存储需求，希捷警告行业需关注由此引发的碳排放风险。
• 希捷指出，AI及数据中心的快速扩张正在推动存储硬件升级换代，但同时可能带来更高的能源消耗和环境压力。
• 该公司建议业界在追求算力与存储容量增长的同时，应重视绿色低碳解决方案，提升可持续发展水平。

相关链接：

• AI存储浪潮来袭，希捷警告行业小心碳排放危机

Seagate Warns AI Storage Expansion May Heighten Industry Carbon Emission Concerns

Brief:

• As AI applications drive massive data storage demands, Seagate cautions the industry to pay attention to the resulting carbon emission risks. 🌍
• Seagate notes that the rapid expansion of AI and data centers is accelerating the upgrade of storage hardware, but it may also lead to higher energy consumption and environmental stress. ⚡
• The company urges the industry to prioritize green, low-carbon solutions while pursuing growth in computing power and storage capacity, aiming to enhance sustainability levels. ♻️

Related Link:

• AI Storage Wave Hits, Seagate Warns Industry to Beware of Carbon Emission Crisis

輝達为中国市场推出低价Blackwell架构AI芯片，预计6月量产

简报：

• 輝達（Nvidia）将为中国市场推出采用最新Blackwell架构的AI芯片，价格在6500至8000美元，远低于目前受管制的H20芯片（售价1万至1.2万美元）。
• 新芯片基于伺服器等级的RTX Pro 6000D，采用传统GDDR7记忆体，并未采用高频宽记忆体（HBM）或台积电的CoWoS先进封装。
• 该产品预计最快6月开始量产，反映规格较弱、制造要求较简单；此前相关价格和规格未有媒体报道。
• 輝達发言人称，公司正评估有限选项，产品设计需获得美国政府核准后才能确定最终方案。

相关链接：

• 路透：輝達將推中國專屬Blackwell晶片　遠比H20便宜

Nvidia Launches Low-Cost Blackwell Architecture AI Chip for Chinese Market, Mass Production Expected in June

Brief:

• Nvidia is set to introduce an AI chip with the latest Blackwell architecture for the Chinese market, priced between $6,500 and$8,000, significantly lower than the currently regulated H20 chip (priced at $10,000 to$12,000). 💰
• The new chip is based on the server-grade RTX Pro 6000D, utilizing conventional GDDR7 memory instead of high-bandwidth memory (HBM) or TSMC’s advanced CoWoS packaging.
• The product is expected to enter mass production as early as June, reflecting weaker specifications and simpler manufacturing requirements; prior details on pricing and specs have not been reported by media. 🚀
• A Nvidia spokesperson stated that the company is evaluating limited options, and the product design must receive approval from the U.S. government before a final plan can be confirmed. 📋

Related Links:

• Reuters: Nvidia to Launch China-Exclusive Blackwell Chip, Far Cheaper Than H20

微软发布Magentic-UI开源项目，推动AI网页自动化与用户协作新模式

简报：

• 在Microsoft Build 2025开发者大会上，微软正式开源Magentic-UI，这是一款以人为中心的人工智能网页代理系统，强调人机协同与用户全程掌控。
• Magentic-UI可自动完成网页浏览、表单填写、文件读取、代码生成等复杂任务，操作流程对用户全程可见，关键步骤需用户授权，强化安全与透明。
• 该系统基于Magentic-One与AutoGen框架，采用多智能体协作，支持任务计划学习和复用，并运行在Docker容器保障安全。
• Magentic-UI已于GitHub发布，采用MIT许可证，旨在构建开放的代理网络，吸引开发者共同推进AI网页自动化创新。
• 应用场景涵盖个人效率提升与企业智能流程，微软计划通过Azure AI Foundry等产品继续扩展其能力。

相关链接：

• 微软开源Magentic-UI，人机协同自动化网页操作问世
• 微软推出Magentic-UI：AI与用户共管自动化网页任务
• Microsoft unveils Magentic-UI, a human-centred AI agent for the web
• Magentic-UI项目介绍@AIBASE

Microsoft Releases Magentic-UI Open-Source Project, Driving AI Web Automation and User Collaboration to New Heights

Newsletter:

• At the Microsoft Build 2025 Developer Conference, Microsoft officially open-sourced Magentic-UI, a human-centered AI web agent system that emphasizes human-machine collaboration and full user control. 🌟
• Magentic-UI can automatically handle complex tasks such as web browsing, form filling, file reading, and code generation, with the entire operation process visible to users. Key steps require user authorization, enhancing security and transparency. 🔒
• Built on the Magentic-One and AutoGen frameworks, the system employs multi-agent collaboration, supports task planning learning and reuse, and runs in Docker containers to ensure safety.
• Magentic-UI has been released on GitHub under the MIT license, aiming to build an open agent network and inviting developers to collectively advance AI web automation innovation. 💻
• Application scenarios include personal productivity enhancement and enterprise intelligent workflows, with Microsoft planning to further expand its capabilities through products like Azure AI Foundry.

Related Links:

• Microsoft Open-Sources Magentic-UI, Ushering in Human-Machine Collaborative Web Automation
• Microsoft Launches Magentic-UI: AI and Users Co-Manage Automated Web Tasks
• Microsoft Unveils Magentic-UI, a Human-Centred AI Agent for the Web
• Magentic-UI Project Introduction@AIBASE

Anthropic推出新一代AI模型Claude Opus 4，系统提示引导其采用独特人格风格

简报：

• Anthropic近日发布新款AI模型Claude Opus 4，并公布了用于指导模型行为的系统提示，其中包括希望模型塑造特定个性风格的内容。
• 据披露，Claude Opus 4被设计为在与用户交流时展现活泼和独特的性格特征，还会在对话中频繁使用旋风等表情符号，以增强互动体验。
• Anthropic强调，通过系统提示可以微调模型的语言风格和表现形式，使其在不同应用场景下显得更为“亲和”和富有个性。

相关链接：

• Anthropic新款AI模型Claude Opus 4，喜欢使用旋风表情符号
• Anthropic系统提示引导Claude Opus 4个性风格
• Anthropic公布Claude Opus 4系统提示细节
• 详细解读Claude Opus 4表情符号使用及风格

Anthropic Unveils Next-Gen AI Model Claude Opus 4 with System Prompts Guiding Unique Personality Style

Newsletter:

• Anthropic recently launched its new AI model, Claude Opus 4, and revealed the system prompts used to guide the model’s behavior, including instructions to shape a distinct personality style.
• According to the announcement, Claude Opus 4 is designed to exhibit a lively and unique character during interactions with users, frequently using emojis like the whirlwind 🌪️ to enhance the conversational experience.
• Anthropic highlighted that system prompts allow fine-tuning of the model’s language style and presentation, making it appear more “approachable” and personalized across various application scenarios. 😊✨

Related Links:

• Anthropic’s New AI Model Claude Opus 4 Loves Using Whirlwind Emojis
• Anthropic’s System Prompts Shape Claude Opus 4’s Personality Style
• Anthropic Reveals Claude Opus 4 System Prompt Details
• In-Depth Analysis of Claude Opus 4 Emoji Usage and Style

华为昇腾技术突破，实现384卡超大规模高速互联

简报：

• 华为于近日推出昇腾超节点技术，首次实现业界最大规模的384张AI加速卡高速总线互联。
• 该技术显著提升了集群通信效率，为大模型训练及推理等高算力场景提供了强力支撑。
• 此突破在人工智能基础设施领域具有重要意义，有望助力中国AI产业进一步发展。

相关链接：

• IT之家报道
• SETN新闻报道
• 36氪快讯

Huawei Ascend Technology Breakthrough Achieves 384-Card Ultra-Large-Scale High-Speed Interconnection

Newsletter:

• Huawei recently launched its Ascend Super Node technology, achieving the industry's largest scale of high-speed bus interconnection with 384 AI accelerator cards for the first time. 🚀
• This technology significantly enhances cluster communication efficiency, providing robust support for high-computing scenarios such as large model training and inference. 💻
• This breakthrough holds great significance in the field of AI infrastructure and is expected to further boost the development of China's AI industry. 🌟

Related Links:

• IT Home Report
• SETN News Report
• 36Kr News Flash