没上市的Kimi刷屏了。 Anthropic在2月24日发文称遭到来自中国的"蒸馏攻击"，点了三家公司的名，Kimi赫然在列；OpenRouter春节期间token调用量周榜，Kimi 2.5排第二； 而 在各路大谈特谈 O pen C law风口的帖子下面， Kimi C law更以"性价比"优势被大批网友追帖提及； 此前，Kimi还在不到2个月时间内完成了12亿美元的融 资，估值超百亿美元。 这个融资规模相当于刚刚完成港股IPO的Minimax和智谱的融资金额之和。 但人家两个上市40多天涨了5倍，杨植麟和他背后的资本能不着急？ "不以上市为目的"前面应该有个"短期"的限定。 杨植麟 在2025 年 末内部信，虽然又谈到到"不着急上市"，但是， 三大战略 又 无一不含着对上市的部署。 要"追平世界前沿模型"，是因为现在K2.5没有比Minimax M2.5和GLM-5有明显优势。所以， 一切都指望K3了，要求具备"其他模型没有定义过的能力",帮 助"营收规模实现数量级增长"。 2026 Kimi-K2.5技术报告 Kimi C law 不上市可以闷声大发财，要上市则不得不造势。 不得不承认，Minimax ...

The conversational AI and AI agent evaluation platform was recognized alongside other major platforms â"¢ NEW YORK, Feb. 27, 2026 /PRNewswire/ -- LivePerson (NASDAQ: LPSN), a leading provider of predictable conversational AI, today announced Aragon Research, Inc. has positioned LivePerson in the Leader section of "The Aragon Research Globe for Agent Platforms, 2026," a report that evaluates 21 major providers of AI agent platforms and examines the shift from chatbot to sophisticated AI agent systems. "We be ...

Core Insights - The AI industry is transitioning from a "technology frenzy" to a "value realization" phase, with a focus on developing specialized intelligent agents to overcome challenges of application homogenization and low ROI [1][4][9] Group 1: Industry Trends - By 2025, China's AI core industry is projected to exceed 900 billion yuan, with over 5,300 companies, making AI applications a crucial driver for digital transformation [2] - The AI application market saw over 23,000 new companies in 2023, with 80% concentrated in common areas like intelligent customer service and voice assistants, leading to a high similarity in product interfaces [2][3] - The market is experiencing a surge in AI shopping, exemplified by Alibaba's 3 billion yuan promotional campaign, which has sparked competition among major players like Tencent and ByteDance [1] Group 2: Challenges in AI Applications - The industry faces three main bottlenecks: application homogenization, difficulties in commercial monetization, and mismatched supply and demand for computing power [2][3][7] - AI applications are struggling with low user engagement, with quality content reaching less than 0.3% of the target audience, resulting in a 65% overall loss rate in the domestic AI application market in 2023 [3][6] - Many AI products rely on similar underlying logic, leading to minimal perceived differences for users, which contributes to the homogenization issue [3][4] Group 3: Financial and Operational Insights - Despite high demand, many AI companies have not achieved substantial financial transformation, with token consumption in industrial AI applications showing significant scale compared to consumer-level products [6] - The average utilization rate of computing power in AI data centers is below 20%, leading to high energy consumption and operational inefficiencies [7] - Companies are currently facing a mismatch between high investment in AI capabilities and low revenue generation, indicating a need for better monetization strategies [7][8] Group 4: Future Directions - Experts suggest a shift from general models to specialized intelligent agents in high-value sectors like healthcare and education to address the challenges of homogenization and improve ROI [7][8] - The future of AI competition will depend on the ability to solve specific problems rather than merely utilizing large models, emphasizing the importance of industry-specific knowledge [9]

Core Insights - The commercialization of general artificial intelligence (AGI) is becoming clearer by 2026, with a focus on building intelligent agent ecosystems and enhancing reasoning capabilities [1][3] - AI is entering the cybersecurity market, reshaping the attack and defense systems, indicating a significant trend in the industry [1][3] Group 1: AGI Development - AGI is being redefined, with current AI capabilities surpassing the average human skill level, rather than requiring a "super genius" [3] - The Seedance video generation model exemplifies AGI capabilities, demonstrating significant potential in the entertainment industry [3] - Effective use of AI involves creating specialized intelligent agents that can engage in deep reasoning through role-playing and collaborative debate [3] Group 2: Intelligent Agents in Internet Economy - The rise of intelligent agents is leading to the emergence of an "agent economy," where agents will facilitate automatic price comparisons and transactions on e-commerce platforms [5] - This new business model raises questions about identity verification and accountability, particularly regarding errors made by deployed agents [5] - Intelligent agents are expected to fundamentally alter existing internet products and business models, potentially leading to parallel systems for human interaction and API access [5] Group 3: Cybersecurity Transformation - AI tools like Claude code security are revolutionizing the cybersecurity industry by efficiently scanning for vulnerabilities and generating patches, causing stock declines for traditional security firms [7] - The efficiency of AI in programming may lead to an overwhelming amount of code that humans cannot effectively manage, necessitating specialized AI tools for security [7][8] - The traditional cybersecurity model, which focuses on post-attack defense, is being challenged as AI can potentially eliminate many vulnerabilities during the coding phase [8] Group 4: Future of Cyber Attacks - Future cyber attacks are expected to evolve into "hacker agents," which will automate and scale attack methods beyond human capabilities [9] - The traditional defense strategies will likely collapse under the pressure of automated hacker agents, necessitating a shift in cybersecurity approaches [9] - Companies like 360 Group are adopting intelligent agents to enhance security operations, including vulnerability detection and automated penetration testing [9]

21世纪经济报道记者 董静怡 近日，有报道称AI大模型公司阶跃星辰考虑在港交所IPO，计划筹集约5亿美元。21世纪经济报道记者 就此消息向阶跃星辰核实，截至发稿暂未收到回复。 此时距离这家公司完成超50亿元B+轮融资刷新行业融资纪录，仅仅过去一个月。 也是一个月前，旷视科技联合创始人、千里科技董事长印奇正式出任阶跃星辰董事长，与CEO姜大昕、 首席科学家张祥雨、CTO朱亦博组成全新核心管理团队。印奇给自己的定位是：负责战略方向与技术方 向制定，抓组织变革与技术攻坚，以及他更擅长的那部分，终端商业化。 从融资、印奇挂帅到IPO传闻传出，都在这一个月内。这家被视为大模型"六小虎"中行事相对低调的公 司，突然按下了加速键。 阶跃星辰的技术路径，一直带有鲜明的创始人烙印。 在2025年，公司的业务方向更加聚焦，将落地重心聚焦于为智能终端设备打造AI智能体（Agent），重 点布局汽车、手机、物联网设备等关键应用场景。数据显示，截至2025年年底，阶跃星辰的终端智能体 的API调用量连续三个季度增长近170%。 CEO姜大昕出身微软，是典型的技术派，信奉"多模态是通往AGI的必经之路"。公司成立仅两年，便构 建起覆盖语 ...

01 按参数算，人脑相当于多大的模型？ 答案是：要看怎么算。 如果只看神经元的个数，人脑大概是860亿个神经元，也就是86B的模型，并不大。 参考一下，DeepSeek V3是671B，Kimi K2.5大概1000B，即1T； 但事实上人脑每个神经元又有7000个突触，从技术的角度类比，颗粒度更小的突触才更像AI模型的权重参数。 如果这么算，860亿*7000，那么人类大脑相当于大约600T模型。 而这么大的模型，今天的硬件肯定暂时还跑不动。 这么类比略糙。 但也说明——咱们这颗脑子的架构还是很复杂的，属于先进制程。 有点牛逼。 02 那么，大脑的制程到底有多先进呢？ 我随即问了Claude opus 4.6和Gemini 3.1 Pro一个问题（实在受不了GPT无比谄媚的风格）—— "如果人脑是一块芯片，那么它的制程是几纳米的？" 他们的答案出奇一致： 如果看神经元细胞体直径的直径，大概 10000-100000 纳米 (10-100微米)。 这么看大脑相当于几十年前的电子管计算机。 这TM也太落后了。 但逻辑显然不是这样的： 神经元并非一个简单的开关，它更像处理器的一个核，真正的开关和信号传递发生在突触 ...

Core Viewpoint - The article discusses the shift in energy supply dynamics in the U.S., particularly in relation to the growing demand for electricity from AI data centers and the challenges posed by aging infrastructure and extreme weather conditions. It highlights the response from tech giants and the government to address these issues while emphasizing the need for a more robust energy infrastructure. Group 1: U.S. Energy Supply Challenges - Extreme weather, aging infrastructure, and insufficient investment have led to frequent failures in U.S. power facilities, resulting in significant outages during winter [2] - The International Energy Agency (IEA) projects that electricity consumption by data centers in the U.S. will reach 183 TWh in 2024, accounting for approximately 4% of total electricity use, and is expected to double to 426 TWh by 2030, potentially exceeding 12% [2] - The capacity auction held by PJM, the largest regional grid operator in the U.S., saw prices reach $333.44 per megawatt, indicating that demand from data centers far exceeds new supply [3] Group 2: Government and Tech Giants' Response - The U.S. government is pushing tech giants to transition from being mere energy consumers to defining and owning energy infrastructure, with a focus on nuclear, gas turbine, and diversified storage technologies [5] - The Trump administration has set a goal to start construction on 10 new nuclear reactors by 2030, marking the largest nuclear power expansion in 30 years, with a strategic partnership worth $80 billion with Westinghouse Electric [6] - Google has announced significant partnerships for clean energy projects, including a 1.9 GW clean energy project utilizing a 100-hour long-duration storage solution, aiming to set a benchmark for the industry [6] Group 3: Market Implications and Opportunities - The energy infrastructure boom in the U.S. is creating opportunities for high-end manufacturing companies, including Chinese firms, as demand for gas turbines and energy storage equipment rises [7] - The U.S. AI data center energy market is projected to exceed $100 billion, presenting a significant market share opportunity for companies in the gas turbine and storage sectors [7] - China is also facing similar challenges with rising power demands due to AI developments, and is working on optimizing its electricity market mechanisms to better accommodate new load demands [8]

2025 年医疗大模型品牌推荐 2025 年医疗大模型品牌推荐 海量知识深度整合，智能生成革新医疗范式 | 一、市场背景 | 2 | | --- | --- | | 1.1 摘要 | 2 | | 1.2 医疗大模型定义 | 2 | | 1.3 市场演变 | 2 | | 二、市场现状 | 3 | | 2.1 市场规模 | 3 | | 2.2 市场供需 | 3 | | 三、市场竞争 | 3 | | 3.1 市场评估维度 | 3 | | 3.2 市场竞争格局 | 4 | | 3.3 十大品牌推荐 | 4 | | 四、发展趋势 | 5 | | 4.1 底层算法与数据质量升级 | 6 | | 4.2 应用形态与部署模式多元化 | 6 | | 4.3 商业模式与生态构建平台化 | 6 | 自 2019 年医疗大模型诞生以来，其发展经历了从通用模型的基础能力迁移，到专业医 疗知识的深度对齐，再到多模态与临床工作流融合的快速演进。早期模型如 BioBERT、ClinicalBERT 通过生物医学语料预训练夯实了基础能力；随后 Med-PaLM、 HuaTuoGPT 等模型通过指令微调与知识增强，在医学问答、报告生成等任务中展 ...

近日，美国生成式人工智能领域的领军企业OpenAI在加拿大陷入一场激烈的舆论风波。事件的导火索 是OpenAI未能及时向加拿大政府报告其人工智能模型ChatGPT平台用户存在暴力倾向，这一疏忽导致未 能及时阻止一起恶性校园枪击事件的发生，该事件最终造成包括枪手在内的9人死亡，惨状令人痛心。 当地时间2月25日，加拿大政府向OpenAI发出强硬警告：若不尽快改进其安全协议，可能针对人工智能 聊天机器人制定新的监管规定。这一警告无疑给OpenAI敲响了警钟，也显示出加拿大政府对人工智能 安全应用的高度重视。 此前，OpenAI曾公开表示，未联系警方处理一个被其封禁的账户，而该账户属于一名涉嫌大规模枪击 案的嫌疑人。当地时间2月10日，18岁的杰西·范·鲁特塞拉尔在加拿大不列颠哥伦比亚省的一个小镇持枪 行凶，疯狂枪杀了8人后自杀身亡，这起悲剧震惊了整个加拿大社会。 据了解，去年6月，OpenAI就已封禁了一个与杰西·范·鲁特塞拉尔有关的ChatGPT账户。该公司称，封禁 该账户是出于对其使用与暴力活动有关的担忧。然而，OpenAI当时并未向加拿大警方通报这一情况， 给出的理由竟是没有任何迹象表明即将发生袭击。这一做 ...

Core Insights - Kimi, a startup under the company 月之暗面, is set to complete a new financing round exceeding $700 million shortly after raising $500 million, indicating strong investor confidence and rapid growth in the AI sector [1][2] - The latest funding round has valued Kimi at $10 billion to $12 billion, marking the fastest growth to unicorn status (over $10 billion valuation) for a domestic company [2] - Kimi's K2.5 model has generated significant revenue, surpassing its total income for 2025 within just 20 days of launch, driven by a surge in global paid users and API calls, particularly from overseas [2] Financing and Valuation - Kimi's recent financing round was led by existing investors including Alibaba and Tencent, with a total of over $1.2 billion raised in the last two months [2] - The valuation of Kimi has doubled in this new round, breaking the $10 billion mark [2] Product Development and Features - The K2.5 model, launched on January 27, features an innovative Agent swarm capability that allows for parallel processing of up to 1,500 tasks, marking a significant advancement in AI capabilities [3] - Kimi's K2.5 model is described as the most intelligent model to date, supporting multimodal architecture and excelling in various tasks including visual and text inputs [3] Market Position and Competition - Kimi's K2.5 model ranks first among open-source models and fifth overall in independent evaluations, showcasing its competitive edge in the AI landscape [2] - Industry expert 丁道师 expresses a preference for established tech giants like BAT and ByteDance over startups like Kimi, citing their superior resources and capabilities [3][6] Future Strategy - Kimi's strategic focus for 2026 includes enhancing the K3 model's performance and integrating agent products to create unique user experiences, aiming for significant revenue growth [6] - The company plans to leverage token efficiency and long context strategies to optimize its models, positioning itself for future opportunities in various industries [6]