前不久，郑州一所中学的语文教师孙成发现，班里五六名学生课间讨论的话题都与最近兴起的 DeepSeek有关。孙成便与他们一同开启了关于使用人工智能（AI）的探讨。 随着讨论的深入，孙成发现，学生对AI的了解、依赖程度，以及AI能够为学生提供的帮助远超自己的 预期：AI如同一个"无所不能"的人，为学生解决各种学习上的问题，甚至会为学生拓展不少课堂上都不 曾涉及的内容。 随着DeepSeek等大模型广泛应用，一股强劲的AI浪潮涌入多个领域，教育也不例外。 接受记者采访时，多名教师都提出了同样的疑问：AI时代，教育该如何发展？ 最近，"小学生手把手教你3分钟速成PPT"的视频引起广泛讨论。视频中一名小学生放学后来到电脑 前，用DeepSeek和Kimi制作了一个关于西安的PPT。对于视频中的操作，不少网友表示"这样以后的小 孩都不用动脑子了，可怕"。 孙成和学校的不少教师都开始使用DeepSeek备课和制作课件。 "AI可以很好地理解教师的问题，比如教师拿不准一些字词的解释和文言文时，可以问问豆包，豆包会 给出解释或建议，但不一定是对的，教师还需要自己思考。"孙成说。 实际上，网上与之相关的话题层出不穷，比如"14岁 ...

Core Viewpoint - SK Hynix emphasizes that the commercialization of the next generation of HBM (High Bandwidth Memory) requires technological advancements across various fields, particularly in power efficiency, and closer collaboration with major foundries is expected [1][3]. Group 1: Development Focus of HBM - SK Hynix's next generation HBM development focuses on three main tasks: bandwidth, power, and capacity [3]. - The bandwidth is a critical measure of data transfer speed, with the number of I/O ports for HBM4 expected to double compared to HBM3E, reaching 2,048 [3]. - Future HBM is anticipated to improve in power consumption and capacity, with the number of DRAM stacks expected to increase from a maximum of 12 to 16 or even 20 layers [3][4]. Group 2: Challenges in HBM Production - To stack more layers of the next generation HBM within a limited height specification of 775 micrometers, the spacing between each DRAM must be reduced [4]. - SK Hynix is advancing hybrid bonding technology, which connects DRAMs directly without bumps, thus reducing chip thickness and improving power efficiency [4][5]. - However, hybrid bonding faces challenges in commercialization due to high technical difficulty and issues with mass production and reliability [5]. Group 3: Competitive Landscape - Samsung has successfully produced 16-layer stacked HBM3 memory using hybrid bonding technology and plans to mass-produce HBM4 using this technology [6][8]. - Micron Technology is on track with its HBM4 development, expecting to start mass production in 2026, with HBM4E following shortly after [12]. - Micron's HBM4 will utilize 1β (5th generation 10nm class) DRAM technology, integrating up to 16 DRAM chips per stack, each providing 32 GB capacity, with a peak bandwidth of 1.64 TB/s [12][13]. Group 4: Future Projections - The HBM4 and HBM4E are seen as crucial for the ongoing expansion of AI performance, with expectations for significant improvements in density and bandwidth [22]. - Nvidia's upcoming AI accelerators are projected to utilize HBM4 technology, with the Rubin Ultra expected to feature up to 1TB of memory capacity [20][22]. - The competitive landscape is intensifying, with both Samsung and SK Hynix planning to adopt advanced foundry processes for HBM4 production, aiming for enhanced performance and efficiency [16][17].

Group 1: Core Innovation of Malu Grape RWA Financing Case - Malu Grape, a landmark agricultural brand in Shanghai, completed a 10 million yuan equity financing through a Real World Assets (RWA) project in 2024, becoming the first data asset securitization case in the agricultural sector [1] - The project combines data assetization with blockchain technology, ensuring transparency and immutability by recording environmental and economic data on the blockchain, thus addressing liquidity and financing challenges in traditional agriculture [2] - A multi-party collaboration involving Shanghai Data Exchange, Left Bank Xinhui, and law firms established a complete chain for data collection, verification, and trading, with smart contracts ensuring compliance throughout the process [3][4] Group 2: Economic Benefits and Industry Upgrade - The financing will be used for smart agricultural facility construction, such as intelligent irrigation and environmental monitoring, expected to generate an additional annual income of 3 million yuan and enhance brand premium [5] - The project promotes a shift from experience-based planting to data-driven precision management, optimizing the supply chain and enhancing product traceability [6] Group 3: RWA Model as a New Financing Paradigm - RWA transforms physical assets into on-chain tokens via blockchain technology, addressing traditional asset pain points such as enhanced liquidity and reduced investment thresholds [7] - The model improves transparency by ensuring asset authenticity and traceability through data on the blockchain, thereby reducing information asymmetry [8] Group 4: Diverse Application Scenarios - The Malu Grape case validates the feasibility of data assetization in agriculture [9] - In the renewable energy sector, Longxin Technology collaborates with Ant Group to provide low-cost financing for charging station operators through RWA [10] - Tokenization can release liquidity in real estate and support emerging asset classes like carbon credits [11] Group 5: Technology Integration and Compliance Pathways - The project integrates blockchain, AI, and IoT, utilizing smart hardware for real-time data collection and smart contracts for automated transactions [12] - A regulatory framework is necessary to balance innovation and risk, with Hong Kong exploring compliance pathways through sandbox mechanisms [12] Group 6: Comparison with Other Financing Methods - RWA offers more flexible asset segmentation and lower transaction costs compared to traditional asset-backed securities (ABS), which rely on credit and have lower liquidity [13] - RWA covers a broader range of asset types compared to security token offerings (STO), combining features of both [14] - The market for RWA assets is projected to reach trillions by 2030, covering sectors like real estate and green energy [15] Group 7: Insights and Challenges - RWA is suitable for enterprises with high-value non-standard assets needing improved liquidity, requiring data governance capabilities and compliance teams [18] - Regulatory uncertainty and technological security are challenges that need to be addressed, including the need for third-party audits to prevent data tampering and smart contract vulnerabilities [20][21]

如果您希望可以时常见面，欢迎标星收藏哦~ 来源：内容 编译自chosun，谢谢。 受通用存储器半导体价格低迷影响，三星电子和SK海力士今年第一季度业绩或将不及预期。业界 预计半导体市场将从第二季度开始全面反弹，但预计复苏将逐步展开，难以出现大幅好转。 据财经信息提供商FnGuide 31日报道，三星电子今年第一季度营业利润预计在5万亿韩元左右，较 去年第一季度（6.6万亿韩元）和去年第四季度（6.5万亿韩元）均减少1万亿韩元以上。业绩下滑 的主要原因是存储半导体价格疲软以及晶圆代工（半导体代工生产）机构部门亏损扩大。 点这里加关注，锁定更多原创内容 *免责声明：文章内容系作者个人观点，半导体芯闻转载仅为了传达一种不同的观点，不代表半导体芯闻对该 观点赞同或支持，如果有任何异议，欢迎联系我们。 10万亿，投向半导体 综合各大券商报告，三星电子DS（半导体）机构部门第一季度内存业务盈利逾2万亿韩元，但预计 营业亏损约4000亿韩元，核心内存业务受到高带宽内存（HBM）出口至中国的管制影响，加上通 用内存产品价格下跌，获利或较去年同期减少。 SK海力士第一季度业绩预计也呈现类似趋势。证券业估计SK海力士第一季度营业 ...

微软共同创办人比尔盖茨（Bill Gates）表示，未来十年人工智能（AI）的发展意味着，世上「多数事情」将不需人类就能完成，但人类在创造力方面， 应有机会维持某种优势。 盖茨2月在美国国家广播公司（NBC）「今夜秀」（The Tonight Show）上，对主持人吉米法伦（Jimmy Fallon）解释，专才目前仍然「稀有」，例如我们 在许多领域当中仍旧依赖的「伟大医生或老师」等人类专家，但「随着AI的发展，杰出的医疗咨询和教学，将在十年内成为免费、常见的服务」。 盖茨上个月接受哈佛大学教授、「快乐」专家布鲁克斯（Arthur Brooks）访问时也说，全球正进入一个「免费智能」的新时代，由AI驱动的技术将快速发 展，变得随处可得，且遍及生活中的几乎每个面向，包含进步的医学与诊断，以及普及的AI教师和虚拟帮手。 —— 我 是 广 告 —— 然而，盖茨重申，特定类型的工作应该永远不会被AI取代，例如，人们应该不会想看机器打棒球。他说，「有些事情我们会保留给自己，但制造东西、 搬运物品和栽种食物等工作，未来基本上都将不成问题」。 盖茨也强调，人类将能决定是否在AI能力所及的范围内，全面使用AI。此外，由于AI无 ...

Core Viewpoint - TSMC's investment in U.S. factories aims to enhance semiconductor production capabilities, but there are significant delays in technology advancement compared to Taiwan, potentially impacting Apple's future chip production [1][2][3] Group 1: TSMC's U.S. Expansion - TSMC has invested billions in its U.S. factories, including a second facility in Arizona set to produce 3nm chips by 2028 and a third facility for 2nm chips expected to be completed by the end of 2030 [1][2] - The production processes in the U.S. will lag behind Taiwan by approximately five years, affecting the availability of advanced chips for Apple [1][2] - Currently, the Arizona factory is producing A16 chips using the N4 process, while Apple will rely on TSMC's Taiwan operations for 2nm chips until the U.S. facilities are operational [2][3] Group 2: Supply Chain and Strategic Implications - Establishing factories in the U.S. helps diversify production and mitigate supply chain disruptions, aligning with Apple's strategy to reduce reliance on Chinese manufacturing [3] - However, this shift may diminish the importance of TSMC's Taiwan operations in the global semiconductor landscape [3] Group 3: Industry Perspectives - Former Intel CEO Pat Gelsinger expressed skepticism about TSMC's ability to restore U.S. leadership in semiconductor manufacturing, emphasizing the need for R&D to be conducted in the U.S. [5][6] - Gelsinger highlighted that TSMC's core R&D will remain in Taiwan, limiting the potential impact of U.S. manufacturing investments [5][6] - He also pointed out that merely investing in manufacturing is insufficient; technological innovation and cost efficiency are crucial for future competitiveness in the semiconductor industry [7]

普通人用AI的八个实践场景：附教程、工具、提示词 从2022年11月ChatGPT发布以来，AI闯入人们的视野已有两年之久，这两年间我的生活因为AI经历了翻天覆地的变化。 在工作中，每当遇到难题或创意枯竭时，我习惯性地向DeepSeek求助，几乎所有问题都能获得满意答案。文章完成后，我会请Claude为我润色一番，不得 不承认，AI那些匠心独运的比喻有时让我叹为观止，可能是我穷尽一生也难以构思出的妙语。阅读外文资料时，受限于自身英文水平，我会借助"沉浸式 翻译"解读全文，或者直接让AI为我翻译全文。 2025年的今天，AI已经深刻改变了我的工作与生活方式。然而令我费解的是，尽管AI发展这么久，各种AI工具如雨后春笋般涌现，但普通人日常生活中 的"AI含量"却始终不高。在与身边朋友交流后，我终于找到了答案。 同事小A：说实话，我到现在都不知道AI能干嘛，我只用过AI帮我算命，你还别说，它算出来说的还蛮有道理的。 朋友小Y：AI这玩意儿太难用了，光是那个什么prompt就搞不明白，一大串话，与其问AI不如直接百度一下来得快。 爸爸老F：啥AI，哎，好，记得常回家，哎。 思索再三，我决定撰写一篇专题文章，分享普通人 ...

本轮港股科技市场行情启动于2024年9月19日，本轮行情起点至今港股科技30ETF（513160）跟踪 的恒生港股通中国科技指数涨幅71.23%，同期恒生科技指数涨57.46%，港股通科技指数涨64.85%，港 股通互联网指数涨57.38%，恒生互联网科技业指数涨43.85%。 华泰证券梳理"阿里巴巴、腾讯、美团、小米、联想、比亚迪、中芯国际"为港股科技"七姐妹"，是 中国科技股估值重塑的重要担当，港股科技30ETF（513160）跟踪的恒生港股通中国科技指数一键布局 港股科技"七姐妹"，合计权重高达60.63%，为港股科技类主题指数中最高，或是布局中国科技资产崛起 的有力指数工具。 港股科技推动AI大模型加速应用！港股科技 30ETF(513160)现涨0.86%，过去20个交易日获得超 5亿元资金净流入 消息面上，3月26日，宝马集团宣布与阿里巴巴集团深化战略合作，基于阿里通义AI大模型，联合 开发AI引擎，将应用于中国市场的宝马新世代系列车型。此次深化战略合作，主要聚焦在AI大模型、 智能语音交互等前沿技术领域。基于通义大模型和斑马元神AI，全新BMW智能个人助理采用宝马与阿 里共同开发的AI引擎，计划 ...

Investment Rating - The investment rating for GaoNeng Environment (603588) is not explicitly stated in the report [1]. Core Views - The report highlights that the environmental engineering segment has negatively impacted the company's performance, while the resource utilization segment has shown steady improvement in profitability [1][3]. - In 2024, the company achieved a revenue of 14.5 billion yuan, a year-on-year increase of 37.04%, but the net profit attributable to shareholders decreased by 4.52% to 482 million yuan due to the decline in the environmental engineering segment and goodwill impairment related to acquisitions [1][3]. - The report anticipates continued growth in revenue and profitability for the resource utilization segment, driven by operational improvements and new project launches [3][6]. Summary by Sections Financial Performance - In 2024, the company reported total revenue of 14.5 billion yuan, with a year-on-year growth of 37.0%. The net profit attributable to shareholders was 482 million yuan, reflecting a decrease of 4.5% [5]. - The gross margin for 2024 was 14.4%, down from 18.2% in 2023, while the return on equity (ROE) was 5.3% [5]. - The company plans to distribute a cash dividend of 4.00 yuan per 10 shares, totaling 609 million yuan, which represents 126.46% of the net profit attributable to shareholders [6]. Segment Performance - The hazardous waste resource utilization segment generated revenue of 11.137 billion yuan in 2024, a significant increase of 72.17%, with a gross margin of 9.14% [3]. - The environmental operation service segment achieved revenue of 1.729 billion yuan, a year-on-year increase of 6.25%, while the environmental engineering segment saw a revenue decline of 34.26% to 1.633 billion yuan [3][6]. Future Projections - The report projects that the company's revenue will reach 16.747 billion yuan in 2025, with a growth rate of 15.5%, and net profit attributable to shareholders is expected to increase to 657 million yuan, reflecting a growth rate of 36.3% [5][6]. - The company is expected to continue focusing on enhancing its operational efficiency and expanding its service offerings in the environmental sector [3][6].

Core Viewpoint - Ant Group has successfully utilized domestic chips, including those from Alibaba and Huawei, in conjunction with the mixed expert (MoE) machine learning method to train AI models, achieving a cost reduction of approximately 20% [1] Group 1 - The performance of the new technology is comparable to NVIDIA's H800 chip [1] - Ant Group continues to use NVIDIA chips for AI development but has shifted its latest models to primarily rely on alternatives from AMD and domestic Chinese chips [1] - Ant Group is continuously optimizing for different chips to reduce AI application costs and has made significant progress, with plans to gradually share its findings through open-source initiatives [1] Group 2 - The move is significant in the context of U.S. export restrictions on high-end chips to China, indicating that China has largely overcome U.S. semiconductor sanctions [1] - Ant Group's open-source Ling series model framework and training strategies could promote the accessibility of domestic AI technology, lowering the entry barriers for small and medium-sized enterprises and research institutions [1]