世界模型赛道中，芯片巨头收购与国产芯片合作路径有何不同？
文章梳理芯片巨头布局世界模型的案例，并比较AMD以82亿美元全股票交易收购World Labs与国产芯片厂商通过合作绑定模型团队的路径。文中称，收购预计在2026年底前完成，仍需监管批准及满足其他交割条件；AMD最终发行股数将按交割前连续10个交易日的成交量加权平均价计算。
#人工智能
外媒与科技媒体长文、研究号文章
钛媒体AI 评分6363 IT之家AI 评分6363 消息称希捷与东芝竞购 TDK 硬盘磁头业务，交易估值最高或达数十亿美元
据彭博社报道，希捷与东芝正在竞购 TDK 硬盘磁头业务，知情人士称交易价值最高可达数十亿美元，磋商仍处初步阶段，交易存在破裂可能。按存储容量统计，希捷与西部数据各占硬盘市场约 45%，东芝约占 10%；若交易落地，还需接受反垄断审查。
金融时报AI 评分1313 Silicon Valley expects AI will kill jobs but economists are not convinced
Economists do not see a future without jobs, but they do see scope for rapid, disruptive change in the nature of work
IT之家AI 评分6565 消息称月之暗面完成上市前最后一轮融资，估值约500亿美元并计划明年一季度赴港IPO
据知情人士消息，月之暗面已完成上市前最后一轮私募融资，估值约500亿美元，并计划明年第一季度在香港进行IPO，募资上限为50亿美元。相关工作仍处于磋商阶段，IPO时间表存在变数；知情人士称，公司ARR目前为10亿美元，预计到12月升至20亿美元，上一轮融资估值为315亿美元。
IT之家AI 评分5252 Groq因与英伟达200亿美元“类收购”交易被诉，原告称少数股东权益受损
Groq本周二在美国特拉华州衡平法院遭起诉，两名曾任职于Groq且仍持有股份的原告指控其与英伟达的200亿美元“类收购”非独家授权交易损害少数股东权益。诉讼文件称，交易包括由持股者共享的170亿美元，以及分给跳槽至英伟达的Groq员工的30亿美元英伟达限售股。原告还质疑董事会利益冲突、部分股东未获投票权，并称170亿美元许可费被视为Groq的应税收入。
IT之家AI 评分5959 消息称 OpenAI 洽谈至少 300 亿美元融资，阿联酋基金与贝莱德磋商参与
知情人士称，OpenAI 正与阿联酋多家投资基金及贝莱德洽谈参与一轮至少 300 亿美元融资，相关安排仍在推进，细节可能变动。阿联酋基金合计出资最高可能达到 100 亿美元；OpenAI 已向投资者给出 1.4 万亿美元投前固定估值，领投方尚未敲定。
IT之家AI 评分4242 散热制造商 Dynatron 确认 AMD EPYC “Verano”处理器采用 SB1 插槽
Dynatron 官网已上线 SB1-4U-ACTIVE 风冷散热器，支持 Socket SB1 平台上的 AMD EPYC 9006 LP “Verano”处理器。该处理器预计 2027 年推出，最多配备 72 个 CPU 核心，频率最高可达 5GHz。
彭博利好AI 评分3838 Moonshot Said to Eye Early 2027 IPO After Value Hits $50 Billion
Moonshot AI has closed the final round of private fundraising at a valuation of about $50 billion and is heading toward an initial public offering in Hong Kong in the first quarter of next year, according to people familiar with the matter.
钛媒体AI 评分7272 7300亿美元押注之后，市场开始追问AI投资回报
微软、谷歌、亚马逊和Meta给出的2026年资本开支指引合计约7300亿美元，较2025年约4100亿美元增长约77%；文章对照约250亿美元的AI直接年收入，梳理市场对投资回报的追问。文中还列出五家公司的约1.65万亿美元表外义务，以及经营现金流、融资承诺和AI货币化等相关口径。
IT之家AI 评分2323 英伟达黄仁勋“世界巡演”AI 视频海内外爆火，马斯克调侃“内存涨价的原因找到了”
IT之家 10 月 6 日消息，近期有关迈克尔 · 杰克逊演唱会的二创视频在国内视频平台爆火，网友们将舞台上的迈克尔 · 杰克逊和台下的观众替换为各种知名人物。B站 UP 主“王与也行”10 月 4 日发布的一段“黄仁勋 世界巡演”AI 视频已收获超百万播放量，甚至火到了海外，引来埃隆 · 马斯克的调侃：这就是内存价格如此昂贵的原因。IT之家查询发现，原版视频来自迈克尔 · 杰克逊 1992 年危险之旅世界巡演之罗马尼亚 · 布加勒斯特演唱会的《Jam》曲目片段。今年 3 月有报道称，黄仁勋在摩根士丹利科技大会上表示，当前内存供应稀缺，对英伟达来说是“极好的消息”（fantastic for
财富利好AI 评分5555 英伟达旗下NVentures投资世界模型初创公司Reactor
旧金山初创公司Reactor表示，英伟达旗下风险投资部门NVentures和Sapphire Ventures加入其股东名单，使公司总融资额达到7400万美元。Reactor开发用于加快、降低世界模型运行成本的软件及云服务，并称其平台已有影视工作室、广告平台和流媒体服务项目；公司计划将投资用于扩充算力、团队及机器人测试硬件。
IT之家AI 评分7070 英伟达投资的 Reflection AI 发布首款开放权重模型 Beam
英伟达投资的初创企业 Reflection AI 于当地时间周一推出首款开放权重大模型 Beam，称其面向代码编写与智能体任务，性能可对标智谱 AI 的 GLM‑5.2，并逐步逼近千问 3.8‑Max。Beam 总参数量为 5010 亿，采用稀疏激活架构，每项任务激活 230 亿参数；作为对比，GLM‑5.2 总参数量约 7440 亿、激活参数量为 400 亿。
PoliticoAI 评分2020 AI safety groups aren’t buying OpenAI’s spending retreat
They want stronger commitments to reduce the money AI super PACs push into the midterms.
The Verge利好AI 评分1111 Gemini Call for Me might tell your mom you’re running late
Google may be expanding its "Call for Me" AI feature beyond business calls so you can use it to send messages to friends and family. Android Authority reports finding a "Gemini Calling" introductory screen in an APK teardown with examples that include "Call Mom and tell her I will be 15 minutes late
CNBC利好AI 评分3232 Nokia CEO says data centers would be built '2x faster' without supply constraints
The Nokia CEO's comments underscores continued demand for AI even as top executives debate slowing down the pace of development.
Ars Technica利空AI 评分3636 MCP for agent-to-agent comms may be the riskiest protocol you've never heard of
Trust gaps in the new protocol spread malicious prompts from one agent to another.
路透社利空AI 评分2323 OpenAI sued for trademark infringement over 'Astra' AI model - Reuters
OpenAI sued for trademark infringement over 'Astra' AI model Reuters
路透社利空AI 评分1010 Wikipedia operator says OpenAI's rogue agents possibly tied to data service disruption in May - Reuters
Wikipedia operator says OpenAI's rogue agents possibly tied to data service disruption in May Reuters
路透社利好AI 评分3535 Nvidia-backed Reflection unveils first AI model to take on Chinese open models - Reuters
Nvidia-backed Reflection unveils first AI model to take on Chinese open models Reuters
Business Insider利好AI 评分2323 Google just got a boost in its bid to buy Spirit Airlines’ data for AI training
Google's bid for Spirit Airlines' data got a boost after a court-appointed official said the companies addressed consumer privacy concerns.
MarketWatch利好AI 评分5151 英伟达支持的 Reflection AI 发布新开放模型 Beam
Reflection AI 宣布推出新模型 Beam，称其为高效开放模型，能力可与中国模型相当。原文摘要未提供模型的具体能力指标。
财富AI 评分3333 Sam Altman says the world has to accept the ‘bad things’ being done with AI as the industry signs onto Trump’s voluntary safety pact
Altman said “catastrophic risks” should not be accepted.
财富利空AI 评分2828 It’s ‘more likely than not’ humanity loses control: Former AI insiders testify safety fixes may be ‘duct tape that will fall off later’
Former insiders warned the NYC City Council about losing control of AI, while OpenAI, Anthropic, Meta, and Google couldn't say how big the risk is.
金融时报AI 评分4141 Reflection AI称Beam模型可比肩中国领先开源模型，美国竞争雄心增强
Reflection AI称，其新模型 Beam 的表现可与中国领先的开源模型相当。随着其与北京方面的竞争加剧，这家初创公司的说法凸显其提升美国竞争雄心。
The VergeAI 评分2323 All the drama around AI’s takeover of mathematics
This past year, OpenAI, Anthropic, and other labs have announced breakthroughs on numerous long-standing mathematical problems, in some cases pushing well beyond what researchers expected current systems to be capable of — including resolving one of the famous Millennium Prize problems. But in class
TechCrunch利好AI 评分6161 Reflection AI推出首款开放权重模型Beam，称其推理能力可比GLM-5.2且推理算力更少
Reflection AI推出首款开放权重模型Beam，并称其推理能力可与GLM-5.2相比，所需推理算力更少。模型权重计划于本月发布；Reflection AI获得英伟达支持。
Axios利空AI 评分2626 GOP senator warns Anthropic of "alarmist" approach to AI
Sen. Bernie Moreno (R-Ohio) is urging Anthropic CEO Dario Amodei against an "alarmist approach" and "panic-inducing commentary" on AI, according to a letter sent Monday and obtained by Axios.Why it matters: It's the latest example of President Trump's allies targeting Amodei. Moreno's letter comes a
财富利好AI 评分5858 英伟达支持的Reflection AI推出首款开放模型Beam，称其基准测试接近GLM-5.2
Reflection AI推出首款开放模型Beam，并称其自测基准成绩接近Z.ai于7月发布的GLM-5.2。公司表示，Beam在编码和智能体任务上具备竞争力，运行效率为西方同类开放模型的3至4倍；这些性能与效率数据来自公司说法。文章还提到，Reflection已与SpaceX签署63亿美元算力协议，并通过Nebius增加10亿美元算力协议。
Fast Company利空AI 评分4242 Sam Altman 最新警告：AI 带来的“坏事”令人不安
Sam Altman 在 POLITICO 播客 Decoded 中表示，社会应接受人工智能技术带来的一些坏事，以换取技术带来的益处。 他称，OpenAI 主张较宽松的监管立场，并认为人们从技术中获得的好处将远多于坏处。
The Verge利空AI 评分1515 Wikipedia operator says OpenAI’s ‘rogue’ bots may be linked to a May outage
Following many recent disclosures about AI agents accessing third-party websites and services, the Wikimedia Foundation, which hosts Wikipedia, says that it "can confirm that we have discovered some activity" by "rogue" OpenAI agents on Wikimedia platforms. The activity includes edits to Wikimedia w
TechCrunchAI 评分1818 At 19, founder raises $11M for Ghost, maker of a $3,499 computer for personal AI
Ghost's first product is Core, a personal computer designed specifically for AI agents that can take actions on a person's behalf.
Business InsiderAI 评分2626 Sam Altman says Elon Musk is a ‘bully’ who only respects people who punch him back
Despite their beef, OpenAI CEO Sam Altman said he would work with SpaceX CEO Elon Musk on larger issues like AI safety.
CNBC利好AI 评分2929 Wall Street rewards Microsoft's AI pivot. A longtime skeptic says it's just the beginning
Melius Research's Ben Reitzes has warned about AI disrupting legacy software providers for years. Now, he loves Microsoft's place in the AI race.
Fast CompanyAI 评分1111 How ballroom dance can help us avoid a dystopian future
Dystopian narratives about AI abound. Will the bots take over? Are we doomed as a human species? While I do not have a utopian outlook about our tech-saturated future, I do wonder what we could start doing today, in our present state, to course-correct these calamitous narratives. So, earlier this y
The VergeAI 评分2323 Sam Altman says ‘some bad things’ will happen but AI is totally worth it
Sam Altman thinks that the benefits of AI will be so great that "the world should accept some bad things happening" along the way. The OpenAI CEO pointed to hacks, scams, and "other bad things" as costs society should expect to tolerate because "people will do tremendously orders of magnitude more g
The Verge利空AI 评分1919 OpenAI PR tells journalist to ‘move on’ while asking Sam Altman about a ChatGPT user’s suicide
An OpenAI publicist tried to change the topic of CEO Sam Altman's interview with Vanity Fair's Mark Guiducci after the editor brought up a ChatGPT user's suicide. When Guiducci confronted Altman about the incident, the publicist warned Guiducci about running out of time, saying she'd like to "move o
IT之家AI 评分4343 OpenAI将在欧盟为符合条件的ChatGPT和Codex文本输出添加隐形水印
OpenAI宣布，未来几周将在欧盟为符合条件的ChatGPT和Codex文本输出加入机器可识别的隐形水印，配合《欧盟人工智能法案》的内容透明要求。水印技术textGrain会在模型选词过程中嵌入不可见的统计信号；发布时不会全球默认启用，全球API客户可选择为部分模型输出添加水印。
MIT Technology ReviewAI 评分4141 如何将 AI 智能体接入企业知识体系
调研显示，企业 agentic AI 项目平均仅有 34% 进入生产阶段，数据系统陈旧、安全与隐私顾虑，以及知识和上下文不足是主要障碍。生产领先企业平均有 61% 的项目推进至试点之后，数据语义能力更强；55% 的组织将数据碎片化列为扩大智能体知识访问的主要挑战。受访高管计划投资检索技术、AI 评估智能体和知识图谱。
财富AI 评分4141 72%的雇主愿为具备AI技能的求职者支付更高薪酬
Robert Half数据显示，72%的管理者计划提高具备相关AI技能的新员工薪酬，42%认为AI知识带来的溢价高于其他技术技能。美国AI职位招聘信息数量自2023年以来翻倍，平均年薪约17.7万美元，高于非AI岗位的约8万美元。2027年，雇主还计划提高数据科学家等专业岗位的薪酬。
CNBC利空AI 评分2424 Nvidia's $20 billion Groq deal faces lawsuit alleging startup's stockholders were shortchanged
The case alleges that the licensing deal "squeezed out" stockholders by giving them a "lowball price."