以下文章来源于AGI接口 ，作者苗正卿 AGI接口 . AI卷起的财富风暴。 出品｜虎嗅科技组 作者｜苗正卿 题图｜视觉中国 OpenAI正处于一个命运十字路口。 对OpenAI而言的好消息是，它大概率即将完成新一轮涉及1000亿美元的公司史上"最大规模"融资；坏消息 是，投资方正在急迫地需要OpenAI证明一个核心命题：它是能够赚钱的。 2月19日，彭博社援引多位匿名知情人士的消息称，OpenAI 正接近敲定一轮创纪录级别的新融资交易第一 阶段，预计该阶段融资能够为OpenAI带来超过1000亿美元资金。在这轮融资后，OpenAI的整体估值可能 会超过8500亿美元。 值得注意的是，这是OpenAI重组为营利性公司后的重大融资，也被视为OpenAI 冲击IPO前的关键一步。此 前消息显示，OpenAI预计在2026年四季度或不晚于2027年内完成IPO。 据悉，1000亿美元的融资，主要会用于OpenAI的 "星门" 超级计算项目以及 GPT-5 系列模型的迭代。根据 此前OpenAI公布的信息显示，"星门"项目预计在2029年前在全球构建10 吉瓦 ( GW )级别的专属 AI 超级计 投资方正在急迫地需 ...

Group 1 - Chinese research team developed an AI medical system called DeepRare [1] - Zhongke Shuguang's SothisAI integrates with Zhipu GLM-5 and Qwen3.5 [1] - Ant Group's subsidiary will release the enterprise version of the Bailing large model [1] Group 2 - Taichu Yuqi completed deep adaptation of Zhipu GLM-5.0 and Alibaba's Qianwen dual open-source models [1] - The national version "OpenClaw" by NetEase Youdao announced open-source, activating the Agent innovation ecosystem [1] - OpenAI's latest funding round is expected to exceed $100 billion [1]

Core Insights - OpenAI is nearing a record fundraising round expected to exceed $100 billion, which would elevate its valuation to over $850 billion [1][2] Fundraising Details - The new funding round, rumored for months, is anticipated to close by the end of the month, with major backers including Amazon (up to $50 billion), SoftBank ($30 billion), Nvidia ($20 billion), and Microsoft [2] - This new round more than doubles the previous record of $40 billion raised by OpenAI in March [3] Financial Strategy - OpenAI is investing heavily in infrastructure, with significant financial commitments to companies like Nvidia and Oracle, raising questions about when these investments will yield profits [3] - CEO Sam Altman emphasized that the current focus is on growth rather than immediate profitability, stating that reasonable unit economics will guide future decisions on profitability [4] IPO Prospects - OpenAI is reportedly preparing for an IPO, potentially as soon as later this year, competing with rival Anthropic to be the first major AI company to go public [4]

Core Insights - GoGo AI Network Inc. announced the successful deployment of an AI-based decision support system by its portfolio company, Algo8 Industrial AI, for a major publicly listed Indian steel and infrastructure group [1] Deployment Scope - The customer operates complex steelmaking facilities that include Electric Arc Furnaces (EAFs), Ladle Refining Furnaces (LRFs), auxiliary steelmaking units, and a dedicated captive power plant, emphasizing the need for efficient energy allocation to manage fluctuating load conditions [3] Solution Overview - Algo8's PlantBrain platform utilizes machine-learning models to analyze real-time power consumption and generate actionable recommendations, enhancing power scheduling and operational resilience [4] - The Power Schedule Mode automates the generation of forward-looking power distribution plans, improving visibility and coordination during unplanned events [4] Strategic Importance - The deployment signifies the increasing adoption of industrial AI solutions in large-scale manufacturing, highlighting Algo8's capability to deliver measurable efficiency gains [5] - The implementation showcases AI's role in optimizing complex energy systems, supporting intelligent and sustainable industrial operations [5] Performance Metrics - Optimization trials indicated average export power savings of approximately 4.39 megawatts (MW) and improved export efficiency, alongside a significant reduction in infirm/unscheduled power without negatively impacting production throughput [7] - Digitalized operational logbooks enhanced workforce productivity by approximately 10.3 labor hours per day, translating to over 3,700 hours annually [7] Operational Optimization - The system recommends start/stop strategies for refining furnaces, predicts optimal ramp-up and ramp-down levels for captive power generation, and improves utilization of import limits while reducing unnecessary power wastage [8]

Core Insights - Agereh Technologies Inc. is positioned for accelerated commercial growth with a strong Board of Directors and executive leadership team as it transitions into its commercial growth phase [1][2] Leadership and Governance - The governance structure of Agereh combines expertise in technology commercialization, public markets, finance, and large-scale enterprise operations, which is crucial for the company's growth [2] - Ken Brizel, CEO and Director, emphasizes the importance of experienced leadership to guide growth and ensure disciplined governance [3] - The Board of Directors includes individuals with extensive backgrounds in technology and finance, such as Jim Plumptre, Mike Plotnikoff, Tim Maddigan, and Rosy Amlani, each bringing over 20 years of experience in their respective fields [4][5][6][7] Executive Management - Ken Brizel leads the strategic direction and product commercialization efforts of Agereh [8] - Joanna L. Hampton, as Chief Financial Officer, oversees financial operations and regulatory reporting, focusing on disciplined governance [9][10] Strategic Growth Initiatives - Agereh's leadership team is focused on expanding its unified intelligent sensing platform across transportation hubs, supporting scalable growth and long-term strategic partnerships [11] - The company aims to build recurring SaaS revenue and enhance its presence in intelligent transportation systems [12] Company Overview - Agereh Technologies Inc. specializes in AI-enabled platforms and sensor solutions aimed at addressing challenges in the transportation industry, enhancing efficiency and optimizing operations [13]

Core Viewpoint - Datavault AI and members of the Nelson Mandela family have announced the formation of Mandela Digital Ventures, aimed at developing digital asset products that align with the legacy of Nelson Mandela [1] Group 1: Company Overview - Datavault AI Inc. is recognized as a leader in data monetization, credentialing, digital engagement, and real-world asset (RWA) tokenization technologies [1] Group 2: Strategic Initiative - The collaboration with Mandela Dlamini & Manaway L.L.C. signifies a strategic initiative to create digital asset products that honor Nelson Mandela's legacy [1]

Core Viewpoint - Alset AI Ventures Inc. is advancing innovation through strategic investment and cloud computing solutions, and has updated its loan transaction with Mr. Randy Gilling for up to $3,000,000 [1] Group 1 - The loan transaction is in accordance with Section 3.5 of TSXV Policy 3.3 [1] - The principal amount of the loan transaction is up to $3,000,000 [1]

硅谷大佬拒绝握手 印度AI峰会"槽点"不断 惠晓霜 林芮竹 正在印度举行的人工智能影响力峰会频现"大型翻车现场"，继有当地大学拿中国产机器狗冒充自主研发 成果引发群嘲后，19日会场又现尴尬一幕：在印度总理莫迪出席的场合，两名与会嘉宾在合影环节拒绝 握手。 【硅谷大佬"不对付" 合影"掉链子"】 两人拒绝握手在社交媒体引发热议，有人评论说："萨姆·奥尔特曼和达里奥·阿莫代伊拒绝握手，却举起 了拳头，这真是尴尬一刻。" 【峰会问题频出 盖茨"临阵"缺席】 峰会16日在新德里揭幕，为期五天，由印度电子和信息技术部主办，号称"有史以来规模最大"。然而， 峰会上风波不断。 19日，峰会展区突然对公众关闭，引发参展企业不满。美国微软公司创始人比尔·盖茨原定当天在峰会 上发表主旨演讲。然而，他的基金会当天早些时候发表声明称，盖茨将不出席峰会。该基金会一名高管 将代表他发言。据路透社报道，盖茨的退出与他被牵扯进爱泼斯坦案有关。 峰会期间的交通混乱亦引发不满。社交媒体上流传的视频显示，由于道路封锁且未安排接驳车，与会者 不得不步行数英里前往会场。印度反对党国大党发言人帕万·凯拉嘲讽说："让工程师们走那么远的 路……同时却在抱怨 ...

自我奖励策略训练模型为什么会导致训练崩溃？ 论文标题： Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models 论文链接： https://openreview.net/forum?id=fDk95XPsCU 代码链接： https://github.com/bigai-ai/LIFT-humanoid Huggingface 链接： https://huggingface.co/collections/TMLR-Group-HF/co-rewarding 本文来自香港浸会大学和上海交通大学的可信机器学习和推理组，已被 ICLR 2026 接收。 目前，RLVR（Reinforcement Learning with Verifiable Rewards）已成为诱导大语言模型推理能力的主流技术路线。然而，RLVR 需要高质量标注数据 来监督奖励获取，这一点是其可扩展性上的主要瓶颈。 一旦走向不需要标注数据的 "自奖励（Self-rewarding）" 强化学习训练，模型往往会迅速陷入训练崩 ...

Group 1 - The core event highlighted a moment of tension between major AI companies, Google and OpenAI, during a public appearance with Indian Prime Minister Modi [2] - During the event, both CEOs, Sundar Pichai and Sam Altman, along with Anthropic's CEO, appeared reluctant to engage in physical contact, opting instead to raise their fists [2] - This incident has been interpreted by some observers as a symbol of a "cold war" in the artificial intelligence sector [2]