Core Viewpoint - Google DeepMind's new model, AlphaGenome, expands AI's predictive capabilities to the complex realm of the human genome, achieving state-of-the-art (SOTA) performance in genomic predictions [1][9]. Group 1: Model Capabilities - AlphaGenome can simultaneously predict 11 different gene regulatory processes, capturing complex interactions within genes [3][11]. - The model accurately analyzes gene splicing mechanisms, identifying how a single gene can produce multiple proteins and when errors occur that lead to diseases [4][8]. - It has demonstrated the ability to predict mutations related to diseases, such as accurately reconstructing pathogenic mutations in the TAL1 gene associated with leukemia [6][23]. Group 2: Performance Metrics - AlphaGenome has achieved SOTA performance in 22 out of 24 evaluations related to genomic trajectory predictions and outperformed existing models in 25 out of 26 direct disease association tasks [14][9]. - The model's predictive performance includes a 49% success rate in identifying regulatory directions for GWAS-related variants, significantly surpassing traditional methods [21]. Group 3: Technical Architecture - The model employs a hybrid architecture combining CNN and Transformer technologies, allowing for high-precision genomic predictions [30][31]. - AlphaGenome's input window extends to 1 million base pairs, enabling it to cover most interactions between remote enhancers and promoters [36]. - The training process utilizes a large-scale dataset covering both human and mouse genomes, ensuring the model learns universal rules of gene regulation across different physiological environments [37][38]. Group 4: Training Strategy - AlphaGenome implements a two-phase training strategy to balance generalization and inference efficiency, including a pre-training phase with strict cross-validation and a distillation phase for model refinement [40][41]. - The training incorporates rigorous data augmentation strategies to enhance the model's robustness against unseen mutations [43].

Core Insights - Google has disrupted the IPIDEA proxy network, which sold the ability to route internet traffic through consumer devices to malicious actors, making it harder for network defenders to detect illicit activities [1][2] Group 1: Disruption Actions - Google took legal action to shut down domains used by the IPIDEA network and shared technical intelligence about its software with law enforcement and research firms [3] - Google ensured that its built-in security protection, Google Play Protect, warns users and removes apps that incorporate IPIDEA's software development kits (SDKs) [3] Group 2: Impact of Disruption - Google believes its actions have significantly degraded IPIDEA's operations, reducing the pool of devices available to proxy operators by millions [4] - The disruption may have downstream impacts across affiliated entities due to shared device pools among proxy operators [4] Group 3: Industry Context - Despite the disruption of IPIDEA, the residential proxy providers industry is rapidly growing [4] - The threat from this industry can be mitigated by raising consumer awareness about apps that offer payment for "unused bandwidth" and encouraging the use of official app stores [5] Group 4: Recommendations - The industry is encouraged to require residential proxy providers to show auditable proof of user consent and for app developers to vet monetization SDKs [5] - Tech platforms are urged to continue sharing intelligence and implementing best practices to combat illicit proxy networks [5]

谷歌宣布，将Gemini 人工智能功能整合至Chrome浏览器，是次引入的Gemini功能包括图像生成工具 Nano Banana、主打个人化应用的Personal Intelligence，以及具代理式功能的Auto Browse，用户只需输 入单一指令，便可指示浏览器在网络上自动完成多项任务。 同时，谷歌亦为Chrome推出全新设计，在 浏览器右侧新增 面板，让Gemini聊天 机器人于用户上网期间保持随时可用状态。 ...

Core Insights - Norway's $2.2 trillion sovereign wealth fund has reduced its stakes in major US tech firms, including Nvidia, Apple, Microsoft, and Alphabet, by the end of 2025 [1][2] Group 1: Investment Adjustments - The fund cut its stake in Nvidia from 1.32% to 1.26% and in Microsoft from 1.35% to 1.26%, while still maintaining these companies among its top five investments [2] - The total number of companies held by the fund decreased to 7,201, having exited over 1,000 companies in the last six months of 2025 [3] Group 2: Portfolio Strategy - The fund's strategy includes simplifying its portfolio, which led to exiting stock markets in Moldova, Iceland, Croatia, and Estonia, while adding investments in Jordan and Panama [3] - Approximately 53% of the fund's investments are concentrated in the US across all asset classes, with significant holdings in US Treasuries, Japanese government bonds, and German bunds [3] Group 3: Geopolitical Considerations - An advisory panel has indicated that the fund needs to enhance its preparedness for increasing geopolitical risks, including tariffs and financial sanctions [4] - The fund faced criticism from US Republican lawmakers for its decision to sell out of Caterpillar Inc. last year [4] Group 4: Fund Overview - Norges Bank Investment Management, which manages the fund, operates under a benchmark index set by Norway's finance ministry, limiting its scope for active investment moves [5] - The fund's investment portfolio includes equities, fixed income, real estate, and renewable infrastructure, all outside of Norway [5]

钛媒体App 1月29日消息，根据最新公布的持股清单，挪威规模达2.2万亿美元的主权财富基金在2025年 下半年减持了美国大型科技股，包括其最大持股英伟达。到年底，负责管理该基金的挪威银行投资管理 公司已经减持了其前四大科技股，其中包括苹果、微软和Alphabet。该基金在6月底至年底期间，将英伟 达股份从其投资组合中1.32%的占比减持至1.26%，同时将其持有的微软的股份从1.35%减持至1.26%。这 两家公司仍然是该基金价值最高的五大投资之一，紧随其后的是Alphabet和亚马逊。苹果位列第二。 （广角观察） | | | June 30 Ownership | Yearend Ownership | | --- | --- | --- | --- | | | | (%) 6月30日所有权 | (%) 年末持股比例 | | Name 姓名 | Industry 行业 | (%) | (%) | | NVIDIA Corp NVIDIA 公 | Technology 技术 | 1.32 | 1.26 | | ID | | | | | Apple Inc 苹果公司 | Technology 技术 | 1.2 ...

近日，但斌执掌的东方港湾海外基金向美国证监会（SEC）提供了2025年四季度的13F文件，其美股的最新持仓情况浮出水面。 截至2025年末，东方港湾海外基金清仓了阿里巴巴、奈飞 （Netflix） 、Coinbase等7只股票，持仓标的锐减至10只。期末管理规模约为13.16亿 美元，折合人民币约91亿元，较去年三季度末的12.92亿美元小幅增长。 去年四季度，但斌的调仓动作颇为显著，持仓进一步向科技巨头集中，尤其是大幅加仓谷歌，一举成为其第一大重仓股。 | Stock ↑ | History | Sector | Shares Held or | Market | % of | Previous % of | Rank | Change in | % | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | | | | Principal | Value | Portfolio | | | | Change | | | | | Amt | | | Portfolio | | Shares | | | GOOG | History | ...

前言 《电报解读》是一款主打时效性和专业性的即时资讯解读产品。侧重于挖掘重要事件的投资价值、分析 产业链公司以及解读重磅政策的要点。即时为用户提供快讯信息对市场影响的投资参考，将信息的价值 用专业的视角、朴素的语言、图文并茂的方式呈现给用户。 ①存储芯片+英伟达，用于DDR的产品已进入小批量测评，芯片元件产品已批量用于英伟达GPU，这家公 司产品获得数据中心、AI-GPU等领域客户认可。 ②端侧AI+光通信+谷歌，这家公司是谷歌Gemini AI端 侧硬件核心芯片供应商，布局FTTR光纤技术，构建多维通信矩阵助力端云互联。 ...

Core Insights - The SAT preparation market is experiencing a significant shift as Google's Gemini AI introduces a free full-length practice test feature, which allows users to generate complete mock exams and receive instant feedback, error analysis, and personalized learning suggestions [1][2] - This change indicates a potential revaluation of standardized tutoring services, as the most scalable and standardized parts of the preparation process are now accessible at a lower cost, impacting the pricing logic of traditional tutoring services [2][5] Group 1: Changes in SAT Preparation - The expensive components of SAT preparation are not just the answers to questions but the entire service chain that includes full-length mock exams, error diagnosis, and learning schedule management [2][4] - Gemini's free SAT feature integrates mock testing, feedback, and content supply into a unified platform, challenging traditional high-cost tutoring models [2][4] Group 2: Impact on the Tutoring Industry - The introduction of free AI tutoring tools is expected to pressure the mid-tier products in the SAT preparation market, such as subscription-based question banks and standardized explanation courses, which are highly standardized and easily replaceable [5][6] - Companies like Chegg are already feeling the impact of AI competition, leading to layoffs and business restructuring as students shift towards free AI tools for learning support [6] Group 3: Educational AI Competition - The competition in educational AI is intensifying, with other platforms like OpenAI and Khan Academy also enhancing their AI offerings to provide more integrated learning support [7] - The shift towards process-oriented learning support indicates that platforms are not just competing on AI capabilities but also on how they engage students in their learning journeys [7] Group 4: Future of Standardized Testing - The introduction of free AI tutors raises questions about how learning outcomes will be reliably measured, as the tools can provide extensive support but do not inherently validate student capabilities [11][12] - The educational system will need to redefine the boundaries of AI usage in testing and teaching, ensuring that certain aspects remain independent of AI tools to maintain the integrity of assessments [11][12]

一串长达百万碱基的非编码 DNA 序列，曾如天书般无从解读，如今其调控功能与变异影响皆可被精准预测；一个与癌症相关的罕见突变，曾让病因成 谜，现在却能通过序列分析直接揭示其激活致病基因的全过程。 这都得益于登上权威科学期刊 Nature最新封面的统一 DNA 序列模型 AlphaGenome，这是 Google DeepMind Alpha 系列 AI 的新成员。 AlphaGenome 是一个能够统一预测长 DNA 序列功能的深度学习模型。它突破了以往技术的诸多限制，能够一次性处理长达 100 万个碱基对的 DNA 序 列，并以极高的分辨率精准预测数千种与基因调控相关的分子信号。 更令人振奋的是，AlphaGenome 不仅能"读懂"基因组，还能在短短一秒内评估出遗传变异对多种生物过程的影响。 论文链接：https://www.nature.com/articles/s41586-025-10014-0 为了推进科学研究，AlphaGenome 现已向全球科研界开放：研究人员可以通过 AlphaGenome API 进行非商业研究，并且可以在 GitHub 上访问模型代码 和权重。 这一突破引发了学界的高度 ...

AI正在重塑人机交互范式——用户从搜索关键词转向直接向聊天框提问。据Gartner预测，到2026年，25%的在线搜索将通 过AI聊天机器人完成。面对资本市场的回报压力，AI厂商亟需证明其商业价值，广告成为最直接的变现路径。 2026年1月，谷歌和OpenAI相继宣布，将在其人工智能工具中推送广告，并在一定范围内开始试水。Meta早在2025年10月 已经公开表示，用户与Meta AI的互动将很快被用来投放个性化广告，并据此修改了隐私协议。 然而，广告的性质却与AI的形象和承诺具有结构性冲突。一个被认为可以超越人类偏见、高效获取信息的"超级大脑"，突 然预装了"推销员"指令，客户将感受到某种不适，甚至是"背叛"。这就像你本来想跟一位专家请教知识，甚至期待一次顿 悟时刻，结果却被套路，买了一款专家能从中抽成的产品。现代人并不惧怕形形色色的营销和广告，但在"求知"中被推 销，却可能严重侵蚀用户对AI刚刚建立起的信任。 更重要的是，广告是否和如何植入AI，将成为各路厂商历史性选择的一个岔路口——AI产品的本质和终极价值是什么？将 广告信息融入大模型之中，究竟是系统性升级还是降级？它对人类生活是有益还是毒害？ 一、谷 ...