←搜索结果

按匹配度 · 共 1129 条结果 清除搜索

星期三

  1. 钛媒体 · AGI

    当世界模型实现实时推理,能给AI应用带来什么创业机会?

    Reactor 正构建面向实时交互式视频与世界模型的推理基础设施,开发者可通过统一 SDK 和 API 以 60FPS 以上帧率、低于 40 毫秒延迟运行模型。该公司累计融资达 7400 万美元,最新一轮由 NVIDIA 与 Sapphire Ventures 追加投资。其平台已汇集多款可实时运行的视频与世界模型,并在好莱坞制片厂、广告平台等媒体娱乐场景推进活跃项目。

    曾担任Apple Vision Pro的技术负责人。Taiuti此前还是Luma AI的联合创始人兼CTO,曾为这家3D与视频生成独角兽搭建了底层基础设施。Bryce Schmidtchen则在Apple干了7-8年,参与构建了visionOS软件。 它们的团队还汇聚了来自Apple、Netflix、Meta、Google、Adobe、Replicate以及Microsoft的工程师与研究人员,在图形学、实时系统、交互式媒体以及扩展AI基础设施方面拥有深厚经验。 在Luma AI的经历,让Alberto Taiuti看到了模型研发与模型投入生产之间的鸿沟:研究团队把全部精力都投入到了模型训练上,这就导致他们没有多余的资源和时间去思考如何在大规模环境下运行这些模型、进行推理。 当两位创始人在Apple Vision Pro团队构建产品时,就会思考:如何让模型在端侧(on-device)的极低功耗限制下优化,如何确保它们能在10毫秒以内的延迟下以交互帧率运行,他们擅长对“把极其庞大的系统塞进极小空间”这种极端工程条件进行优化。 过往经历,让他们意识到,当多模态模型及应用的需求爆发,需要一个

    匹配字段正文

  2. Tech.eu

    Google 向 200 名欧洲科学家开放前沿模型并配 300 万美元算力额度

    Google DeepMind 与 Google Cloud 将向最多 200 名欧洲顶尖科研人员提供前沿模型的直接访问权限,并配套 300 万美元算力额度。研究人员可使用 AlphaFold 3、WeatherNext、AlphaGenome、AlphaEvolve 和 AlphaEarth Foundations 等科学工具,Google 称此举意在加快欧洲创新速度与科研竞争力。Google 未说明具体哪些研究人员会获得访问权限。

    models. Several London-based AI labs, founded by former DeepMind scientists, have come out of stealth this year. Ineffable Intelligence, founded by former DeepMind bigwig David Silver, raised $1.1 billion in seed funding, the largest ever in Europe, at a valuation of $5.1 billion, while Inherent, founded by former DeepMind, Microsoft and White House staff, also came out of stealth, emerging with a $50m raise. Gogole; Google Cloud Would you like to write the first comment? Login to post comments

    匹配字段正文

  3. The New Stack

    Chainguard 发布 AI 安全工具 Harness,用 Claude Mythos 5.1 挖出 vLLM 漏洞

    Chainguard 发布 AI 安全工具 Harness,基于 Anthropic 受限开放的 Claude Mythos 5.1 在应用代码中寻找并修复常规扫描器漏掉的漏洞,并计划逐步支持其他前沿模型。Harness 首个公开成果是 vLLM v0.31.0 之前版本的一个漏洞,同一引擎上的一个客户端可篡改其他客户端收到的输出且不写日志,已修复并记录为 GHSA-6cxc-2vcg-w5qc。该工具要求漏洞在隔离环境中可复现才算数,修复方案需通过同一验证并由工程师审核,目前还发现若干关键基础设施软件问题,正在协调披露。

    s in seconds. Agents can run tests, inspect failures, and revise code before opening a pull request, avoiding the CI queues that become harder to manage as vendors like Cognition, Microsoft, and Augment Code turn coding agents into shared team infrastructure running multiple sessions in parallel. To avoid rebuilding environments on every run, Checks uses warm snapshots with dependencies already installed and refreshes them in the background when the lockfile changes. Each workflow step runs in a

    匹配字段正文

  4. Analytics India Magazine

    OpenAI 一名沙特籍工程师因在会议室安装隐蔽录音设备被捕

    美国司法部宣布,OpenAI 工程师 Abdulrahman Alfozan 因涉嫌在旧金山办公室高管会议室安装隐蔽录音设备被捕并被起诉,该设备于 9 月 30 日被发现藏于卫生纸中并接在电视后的 USB 接口上,内容已加密。OpenAI 表示没有迹象显示他不当获取或导出模型权重、源代码或客户数据,公司已将此事移交 FBI 并终止其雇佣关系。起诉书还提到,9 月 24 日至 30 日期间,一个与其配偶关联的聊天机器人账号提交了关于搭建、定制和远程访问录音设备的提示词;若罪名成立,Alfozan 面临最高五年监禁和 25 万美元罚款。

    companies, AI infrastructure startups and venture capital before joining OpenAI. According to his professional profile and media reports, he has held engineering roles at Meta and Microsoft, and worked at Nash, a logistics technology startup focused on delivery infrastructure. His reported background includes a computer science and engineering degree from the Massachusetts Institute of Technology (MIT) and a master’s degree in data science from the University of California, Berkeley. A February

    匹配字段正文

  5. Computerworld

    Mozilla 推出 Firefox Enterprise,进军企业浏览器市场

    Mozilla 发布 Firefox Enterprise,将 Firefox 浏览器与可部署在企业自有基础设施上的管理控制台结合,支持策略、访问与 AI 使用管控,并满足 GDPR、DORA、NIS2、ISO27001、SOC2 等合规要求。该产品基于 Mozilla 自研的 Gecko 引擎,目前已在欧洲的法国、德国、奥地利和瑞士率先推出,后续将扩展更多市场;Mozilla 同时还在开发 Thunderbird Enterprise。

    rol of AI features, and even disengagement if they conflict with policy. Key differentiators Firefox Enterprise isn’t the only option when choosing an enterprise browser, however. Microsoft Edge for Business would likely still be the preferred choice for environments already heavily invested in Microsoft’s enterprise Azure and Microsoft 365 ecosystem, Levy pointed out. Similar logic would apply to Google Chrome Enterprise’s appeal for those already committed to Google’s cloud and productivity in

    匹配字段正文

  6. 钛媒体 · AGI

    所有App都在塞AI,但用户已经烦了

    QuestMobile 数据显示,2025 全年内嵌 AI 应用插件的月人均使用频次同比增速是独立 AI App 的 2.8 倍,月人均使用时长增速是 3.3 倍,但越来越多用户开始反感 App 强行塞入 AI。产品经理认为企业做 AI 更多出于对掉队的恐惧,接入门槛和成本已大幅下降,问题在于如何做、做多深以及什么时候做。

    中的一种,也的确已经给某些产品带来了改变。 应运而生的,是对掉队的恐惧。每个公司、每款应用都无法超脱于这场“军备竞赛”。“当市面上几乎所有产品都在叠加AI功能,如果自己不引入AI,就意味着会损失一部分用户的黏性、体验时长和未来的商业机会。”杨康说。 例如,当Notion推出人工智能写作工具后,Google Docs(谷歌文档)、Evernote(印象笔记)和Microsoft Word(微软Word)等传统软件也不得不进行改革,没有人希望自己的产品被认为是“过时的”“落后的”。 图源:《隐藏人物》 Seany也观察到,现在很多公司的AI焦虑,会逐级传导到每一个员工身上,“比如你作为某一个App的产品负责人,每天都在群里看到老板问,这个AI怎么样了,那个AI怎么样了,为什么我们的产品里还没有体现AI能力?那你就不得不做。”因此,这也让市场上存在“为了有AI而设计AI”的情况。 去年这个时候,调用GPT-4级别模型,每1000个token仍需要几分钱;如今,具备同等能力的模型价格已经下降至原来的1/3以下,国产对标模型的价格则更低。接入AI的门槛也越来越低,企业不需要自研大模型,直接调用A

    匹配字段正文

  7. Thurrott

    XBOX CEO Asha Sharma 宣布成立 XBOX LABS 探索 AI 游戏创作

    XBOX CEO Asha Sharma 在内部邮件中宣布成立 XBOX LABS,由 Scott 牵头组建一支由游戏制作者、工程师和科学家组成的小团队,把新兴想法转化为玩家和开发者可体验的产品,并以更多人游玩和创作作为成功标准。她提到近期游戏被快速反编译并移植到原本不支持的硬件上,Windows PC 版本被转译到 Apple Silicon 或通过 WebAssembly 和 WebGPU 在浏览器中运行,后者全球浏览器支持率已接近 90%。她表示 XBOX LABS 不会改变现有优先级,而是提供探索新技术的途径,并提到人们每年玩游戏超过 720 亿小时,是写代码时间的五倍。

    of another.” It’s fair to say we live in interesting times. Just the past week, ArtCraft announced vibe-coded replacements for the top seven Adobe Creative Cloud apps, and it has Microsoft Office and other top-tier app replacements on the way. And Google announced a Playground vibe-coding tool for creating video games and opened up its Google AI Studio vibe-coding tool for Android and web apps to non-developers. Vibe coding is suddenly a way to make apps–including games–of all kinds. And the me

    匹配字段正文

  8. The Next Platform

    微软公布 Clea Maia 300 推理 XPU 规格

    微软在 Open Compute Project Summit 2026 上公布第三代自研推理加速器 Clea Maia 300 的规格,采用台积电 2 纳米 N2P 工艺和 chiplet 架构,封装内含 7 颗芯片与 8 个 HBM4 堆栈,配备 288 GB HBM4 和 18 TB/sec 内存带宽。微软称一机架 96 颗 Maia 300 的 token 吞吐量最高可达一机架 48 颗 Maia 200 的 7 倍,每瓦性能提升 2.5 倍,单机架功耗约 167 千瓦。该芯片面向更大规模的智能体推理负载,ATL 2.0 网络可将 128 个机架、12,288 颗加速器连入同一一致性域。

    If everyone has an AI inference problem, with tokens still being too expensive to generate even with state of the art GPU and XPU accelerators, Microsoft probably has the biggest inference problem in the world. For one thing, running inference for OpenAI ChatGPT models as well as helping the AI model builder train its models is a big job. And then all of Microsoft’s worker productivity and back office applications are being infused with AI routines, and this is driving usage up the hockey stick

    匹配字段正文

  9. BBC Technology

    人类还需要做翻译多久?

    Google Translate 每月翻译量已达 1 万亿词,机器翻译直销去年估值约 14 亿美元。OKA 研究显示,AI 生成译文中 27% 在发布前仍被修改,其中四分之三是 AI 引入的错误,5% 至 6% 改变了原意。法律和医疗领域的译者指出,AI 会把“Le demandeur”(原告)译成“the requester”,或把希腊字母 μ 误作 milli,导致剂量差 1000 倍。

    higher stakes than the weather. In April, Google said people are now translating, external one trillion words a month with Google Translate. That doesn't count the others: DeepL, Microsoft, Claude, ChatGPT and Gemini, not to mention made-to-measure corporate translation systems. Direct sales of machine translation, external (i.e. outside general AI subscriptions) were worth an estimated $1.4bn (£1bn) last year. AI translation is the "right call" in many cases says John Zimmerman "For plenty of

    匹配字段正文

  10. Fortune AI

    Centene CEO:医疗 AI 最大的机会不在自动化,而在改变护理方式

    Centene CEO Sarah London 在 Fortune Most Powerful Women Summit 上表示,当前大规模部署的医疗 AI 多属“通胀型用例”,只加快计费、自动化文书,未必降低护理成本。她认为更大机会在于把护理下沉到社区、让患者更多掌控自身护理,从而真正压低成本曲线。Centene 在 29 个州运营 Medicaid 计划,服务 2600 万会员。

    ugh its sale, and held senior finance roles at DigitalOcean, where he helped the company navigate its IPO. He also held finance roles supporting the cloud businesses at Amazon and Microsoft. Anne Giviskos was appointed CFO of MoonPay, a financial technology company Giviskos most recently served as interim CFO and chief accounting officer at Navan, an AI-powered business travel and expense platform. She previously held executive roles at the New York Stock Exchange and Euronext. Big Deal Nearly e

    匹配字段正文

  11. CNET

    Google 发布 Gemini 4 Argon,面向编码、数学与机器人

    Google 发布 Gemini 4 Argon,并称已在内部用于软件工程、网络安全与研究。据其博客,Gemini 4 被用于优化 Python 代码库、发现并修补安全漏洞,还让开源 Gemma 模型的任务速度提升三倍,并参与机器人研究,帮助人形设备理解现实物理环境。Google 表示 Gemini 4 通过了白宫新模型自愿审查流程中的安全框架检查。

    Google’s AI Image Generator Too Good? Pixel 11 vs. Pixel 11 Pro vs. Pixel 11 Pro XL: Google’s Latest Phones Compared The Latest from Katelyn Can Vibe Coded Apps Replace Adobe and Microsoft? These AI Apps Hope So; You Can Use Gemini 4 Argon Now: Google’s New AI Is Built for Coding, Math and Robotics; How Microsoft Is Trying to Keep Your AI Agents Contained Google’s Demis Hassabis Puts a Hopeful Spin on AI’s Environmental Impact I Put the Oppo Find X10 Pro’s Camera to the Test. These Are the Resu

    匹配字段正文

  12. Databricks Blog

    Databricks 首次入选 Gartner 分析与 BI 平台魔力象限,获评 Visionary

    Databricks 首次进入 Gartner 分析与商业智能(ABI)平台魔力象限,获评 Visionary,并在所有受评厂商中取得最高的“Completeness of Vision”评分之一,成为该报告 20 多年历史上首发表现最好的厂商。Databricks 将这一成绩归因于数据与 AI 平台融合、智能体分析以及以语义为核心三大趋势,并称 Genie 与 AI/BI 让智能体直接基于 lakehouse 中完整、受治理的数据工作。

    image1.png

    key trends, understand what changed and why, and decide what to do about it instantly. By bringing insights directly to where you work—whether that is through mobile apps, Slack, Microsoft Teams, or custom AI assistants—we are truly putting data in the hands of everyone in the company. We are grateful to the thousands of customers who have partnered with us to push the limits of what is possible with their data. Being named a Visionary with the highest debut performance in the history of the Ga

    匹配字段正文

  13. Ed Zitron 长文 / Where’s Your Ed At

    Ed Zitron 长文质疑 AI 公司年化收入与 GPU 租赁数据

    Ed Zitron 在长文中指出,OpenAI 与 Anthropic 等公司使用的年化收入(annualized run rate)缺乏明确时间区间和计算口径,无法反映真实营收。他引用公开合同与财报称,AI 初创公司以低于成本的价格转售 token,其收入高度依赖持续融资,而 GPU 长期租赁的实际价格也远低于公开指数报价。

    consumption-based revenue, plus recurring monthly subscription revenue, they added. OpenAI, meanwhile, counts only its own cut of proceeds from arrangements with partners such as Microsoft. The $70bn figure reported in September arose from attempts by investors to compare Anthropic and OpenAI performance on a like-for-like basis, according to a person close to OpenAI. However, OpenAI described the $70bn as a “net” revenue metric when it was approached by the FT at that time. Back in August, Ant

    匹配字段正文

星期一

  1. 智东西

    vLLM Semantic Router 开源 Decision 3.0 多模态决策模型系列

    vLLM 社区孵化的 vLLM Semantic Router 推出并开源 Decision 3.0 系列多模态决策模型,共 d3、d3-flash、d3-mini、d3-nano、d3-lite、d3-edge 六款,参数规模覆盖 0.6B 至 27B,权重已在 Hugging Face 开源。旗舰 d3 基于 Qwen3.8-27B 打造,支持输入多张图片和视频,其余 5 款基于不同参数规模的 Qwen3.5。在 Jev Decision Index 0.3.1 榜单中,d3 文本决策指数 64.7 分、公开集 65.5 分位列文本榜第一,超过 Jev 的 60.1 分,视觉榜得分 71.6 分同样第一。团队数据显示,单张 AMD Instinct MI325X GPU 上 d3 处理文本请求延迟中位数 55 毫秒,含图片请求 273 毫秒,10 秒视频请求 553 毫秒。

    源Decision 3.0(图源:X) 这类不生成长文本,专门“做判断”的模型,近期因Jev走红而受到关注。Jev模型上线仅3天,其Token用量便达到1万亿;上线仅24天,其开发商TypeSafe AI便宣布获得a16z领投的8.7亿美元(约合人民币58亿元)融资,估值达到75亿美元(约合人民币502亿元)。 除Jev外,10月10日,微软也推出其决策模型Microsoft-Decision-1,通过Microsoft Foundry和OpenRouter提供服务。该模型基于Qwen3.5-9B微调,对预设选项评分、返回结构化判断及置信度,可用于工具选择、模型路由和Agent流程控制。 vLLM-SR的决策模型Decision 3.0共包含d3、d3-flash、d3-mini、d3-nano、d3-lite和d3-edge六款模型,参数规模覆盖0.6B至27B,相关模型权重已在Hugging Face开源。 开源地址: 其中,旗舰模型d3基于Qwen3.8-27B打造,是一款27B的多模态模型,支持输入多张图片和视频。其余5款模型则基于不同参数规模的Qwen3.5打造。 与Jev不

    匹配字段正文

星期二

  1. The New Stack

    Workday 用 TypeSafe 的 Jev 实时生成界面,Sana 成为默认首页

    Workday 将去年以 11 亿美元收购的 AI 助手 Sana 设为 HR、财务和 IT 软件的新首页,界面由 TypeSafe 的决策模型 Jev 在运行时从设计系统组件中挑选拼装,而非由模型生成完整界面。Workday 首席 AI 官 Joel Hellermark 称,用分类模型从现成组件中选择可以做到即时响应,TypeSafe 称 Jev 的响应时间为 70 至 500 毫秒。Jev 只返回最多 255 个选项中的一个、分数或是否答案并附带概率,不生成文本;TypeSafe 于 9 月 15 日发布 Jev,并在上周宣布由 Andreessen Horowitz 领投的 8.7 亿美元 A 轮融资,估值 75 亿美元。

    cruiting marketplace. There is no dearth of competition in this field, of course. OpenAI , Amazon , Cloudflare, and Perplexity have already shipped their decision models too, with Microsoft following on last week. A pattern developers are already testing Using Jev and similar models for building user interfaces isn’t completely new, though. CopilotKit, an open source framework for adding AI copilots to apps, published a recipe in which Jev picks which prebuilt panel to show, with an OpenAI model

    匹配字段正文

  2. The New Stack

    Workday 将 Trusted Agent Platform 无头化,向外部 AI 智能体开放 HR 与财务数据

    Workday 在拉斯维加斯 Workday Rising 大会上宣布其 Trusted Agent Platform 实现无头化,MCP 服务器、对实时 Workday 数据的 SQL 访问和智能体共享运行时正式可用,客户可在 Anthropic、Google、Meta、Mistral、Nvidia 和 OpenAI 的模型上运行自有智能体。其 MCP 服务器 Agent-Ready Tools 随 Workday Extend AI 提供,目前有 275 个读取工具和 54 个写入或删除工具,用户以本人 Workday 凭证登录,智能体无法访问其账号权限之外的数据。Workday 称过去四个月客户已用 Developer Agent 和 Agent-Ready Tools 构建了 500 多个智能体和 6000 多个应用。

    HR and finance data through copies in a data lake or through service accounts, but, the company argues, those approaches bypass those access controls (though companies like SAP , Microsoft, and others would surely argue that they, too, have access control in place). MCP tools, SQL, and a shared runtime For AI developers, the main entry point to the Workday ecosystem is Agent-Ready Tools. That’s the company’s MCP server, which ships as part of Workday Extend AI. Users sign in with their own Work

    匹配字段正文

  3. SiliconANGLE

    Workday 将 Sana 设为 AI 驱动 HR 与财务的新主页

    Workday 在 Workday Rising 大会上宣布,智能体体验 Sana 成为新的 Workday Home,员工可直接通过 Sana 完成工作,并推出可处理入职、休假和密码请求的 Sana Concierge。Sana 还能借助 TypeSafe AI 的 Jev 模型按员工角色生成表单、日历等交互应用。Workday 同时发布 Work Intelligence 与 Decision Intelligence 两个平台,并同意收购智能体身份平台 C1.ai(ConductorOne)。

    rovides a place to see and manage agents. It will go beyond registering agents to also govern what they can access across Workday and other enterprise systems. Image: SiliconANGLE/Microsoft Designer Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities. 15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more; 11

    匹配字段正文

  4. SiliconANGLE

    Cisco 的 OCP 信号:网络已成 AI 关键路径,开放是入场券

    Cisco 在 OCP 全球峰会上发布首款面向 400-gigabit-per-lane 信令的可插拔光模块 USFP,单模块带宽达 12.8 Tbps,约为现有主流模块带宽密度的四倍,并与 TE Connectivity、Amphenol 推动多源协议标准化。Cisco 同时推出首批 Silicon One G300 系统、G302 机架顶部芯片和带外管理平台,并将 Live Protect 运行时安全能力引入 SONiC。

    mature neoclouds are now closely examining SONiC. Cisco is betting it can win on silicon, optics, systems engineering and support while letting customers choose what runs on top. Microsoft running Silicon One P200 systems in production and Alibaba deploying its in-house router on the P200 at massive scale suggest that bet is paying off. Being open doesn’t hurt Cisco; it expands its addressable market. A Cisco that requires its full stack can sell only to customers that want it. A Cisco that is

    匹配字段正文

  5. Computerworld

    Workday 发布基于 Sana 的默认首页与四个自动执行智能体

    Workday 在 Workday Rising 大会上发布基于 Sana 的默认首页 Workday Home,把任务、决策、会议和智能体活动集中到一处,并可按角色实时生成表单、日历和审批卡片。同时推出 Sana Concierge 员工服务入口,以及招聘、薪酬、会计、合规四个自动执行智能体,可在用户离线时按心跳运行。Workday Home 已可用,Sana Concierge 面向早期采用客户开放,预计 2027 年第一季度全面可用。

    as financial closing or preparation of workforce plans by HR. Sana also works within other AI tools and platforms including Anthropic Claude, Google Cloud’s Gemini Enterprise, and Microsoft Copilot, as well as within Microsoft Teams and Slack. But, Workday said in its announcement, “third-party AI tools and platforms never get raw access to Workday data — they can now access this data through Sana, and Sana does the reasoning, applies each company’s own permissions, approvals, and policies, and

    匹配字段正文

  6. Fortune AI

    Hachette CEO 警告:AI 走捷径正在造就“更差的 C-suite

    Hachette CEO David Shelley 警告,过度依赖 AI 写邮件、做摘要会让高管出现沟通危机,并“钝化大脑”,最终造就“更差的 C-suite”。General Assembly 与 Ezra 调查显示,68% 的英美领导者用 AI 总结文档,58% 用它起草邮件;2025 年 MIT Media Lab 研究发现,用 ChatGPT 写文章的一组大脑参与度最弱。

    o-thirds (68%) of leaders in the U.S. and U.K. turn to AI to summarize documents, and 58% use it to draft emails, according to a survey by General Assembly and coaching firm Ezra. Microsoft executive VP Ryan Roslansky uses AI to help write almost every “high-stakes” email he sends. Microsoft CEO Satya Nadella relies on AI agents to digest his inbox and prepare him for meetings. Meanwhile, Meta’s Mark Zuckerberg is reportedly building an AI clone of himself that is trained on his tone and manneri

    匹配字段正文