搜索“Google”

s are eager to pass the buck to the victims. At an event this week hosted by the bug-hunting software developer Snyk, panelists from the startup, one of its investors, OpenAI, and Google all encouraged organizations to use AI agents to shore up their defenses against, well, other AI agents. “We've got to reinvent everything to match machine scale with machine scale,” said Ed Sim, a general partner at Boldstart Ventures. He and others expressed little interest in regulating AI labs. Manoj Nair, c

匹配字段正文

e your private conversations, address, personal history, and other sensitive data with Meta’s virtual assistant, or one of its competitors. “If I am not storing my stuff in Gmail, Google doesn’t have the plaintext of all the emails I’ve ever sent. But three-quarters of the emails I’m sending are being sent to Gmail accounts,” explained Galperin. In other words, if anyone you’re communicating with uses data-hungry tech platforms or has bad security practices, they can expose everyone’s communicat

匹配字段正文

conomics lecture, but in the pronouncement of a world leader defending his country. conducting a war. Three of Ukraine’s strikes hit data centers that belonged to Yandex, Russia’s Google-equivalent, and damaged the company’s supercomputers that housed some of its AI models. The latest attack hit a data center operated by Sberbank, Russia’s largest bank. It’s too early to tell what the effects of the strike on the Sberbank data center will be, but the three strikes on Yandex have disrupted intern

匹配字段正文

AI最新财务数据引发的争议集中在收入统计方式。 10月8日，据FT，OpenAI向投资者披露，截至9月底，公司年化收入接近500亿美元。但市场此前曾广泛估计，OpenAI的年化收入接近700亿美元，换句话说，这也是“不及预期”。 需要指出的是，两者之间的200亿美元差距，来自OpenAI与Anthropic不同的收入确认方式。Anthropic将通过AWS、Google Cloud等云合作伙伴实现的销售收入计入自身收入。OpenAI没有采用完全相同的处理方式。此前投资者为了比较两家公司，调整了OpenAI的收入数字，形成了700亿美元的市场预期。 从上述口径来看，或许收入数据变化并不代表OpenAI突然损失了200亿美元收入。公司向投资者披露的收入仍在增长，增幅超过70%。 但资本市场对这条消息的反应较为直接。 10月8日，英伟达（NVDA.US)、AMD（AMD.US)、博通（AVGO.US）等主要芯片股集体下跌。纳斯达克指数当天下跌约1.3%，费城半导体指数跌幅超过3%。 这种反应与OpenAI在AI产业链中的位置有关。过去几年，OpenAI是全球AI算力需求增长的重要推动者——公

匹配字段正文

ility and openness, and mandating monitorability would ultimately do the whole industry a favor — it would make all their statements more believable. One report just came out that Google’s newly released Gemini 4 Argon model has a recurrent architecture as well. This is unconfirmed, but Google itself admits Argon outputs less monitorable chains of thought, can control its own chain of thought well enough to say “hello evaluators reading my chain-of-thought,” and, when instructed to, can evade cu

匹配字段正文

，电影早期只是手工作坊里的奢侈品。 AI视频行业里，中国模型可以说包揽了第一梯队的绝大部分席位。在Artificial Analysis的视频竞技场榜单上，Seedance、Vidu、Kling等中国模型多次位居前列。如今，排行榜上位列前茅的多数模型，都来自中国公司。 不仅性能领先，价格梯队上，海外头部模型的API价格仍普遍高于国内模型。 对比头部模型价格，Google的Veo 3.1 Fast含音频约0.15美元每秒，完整版约0.4美元每秒；即便是3月底发布的轻量版Veo 3.1 Lite（720P，0.05美元每秒），价格也只能和国内中高端模型打平，但画质和能力差距明显。 但从下半年开始，在视频模型消耗结构上，轻量模型正在成为走量主力。以OpenRouter的token消耗排行作为参考，可以看到诸如Veo 3.1 Lite、Seedance 2.0 Fast等模型的消耗，超过了其它主流模型。 迭代快、场景深、价格狠——中国模型的“三板斧”砍下去，海外模型很难接得住。 2022年，扩散模型与Transformer的融合刚刚打开视觉生成的天花板。海外，Stability AI开源 St

匹配字段正文

use Wonder翻车了。 结果Muse Wonder交上来的作品既没有玻璃的通透感，里面的熔岩也动不起来，看着就像个随手糊弄的普通台灯。 这么低级的失误直接引来了马斯克围观，他在X上留了一句意味深长的“Hmm”。 花了数百亿美元，Meta难道只练出了一个擅长应试的“做题家”？ Meta AI需要一场翻身仗 过去几年，夹在OpenAI、Anthropic和Google DeepMind中间的Meta，处境一直挺尴尬。 手里握着全球最大的社交平台矩阵，也有Llama开源模型攒下的开发者生态，偏偏在最核心的闭源顶尖大模型牌桌上，Meta始终慢人一步。眼看GPT和Claude轮番刷新能力上限，扎克伯格只能憋屈地扮演追赶者。 小扎心里苦呀。 扎克伯格明显坐不住了。为了改变这一局面，在过去一年里他除了砸重金囤积算力，Meta还专门拉起了一支超级智能团队，把Scale AI创始人Alexandr Wang招致麾下，又从OpenAI和Google DeepMind挖来大批核心骨干。 Muse Wonder就是这场豪赌砸出来的阶段性成果。平心而论，这次新模型倒也真有两把刷子，在不少高难度任务里摸到了

匹配字段正文

ng the best period possible. I am not going to spend much time dwelling on gross versus net. It appears that, per Reuters, 47% of Anthropic’s 2025 revenues came from sales made by Google and Amazon, who take a 16% cut as opposed to the roughly 20% that Microsoft takes from OpenAI. It appears that Anthropic for whatever reason counts every dollar of sales as part of its annualized run rate figure and OpenAI does not, for reasons that escape me, as I simply don’t buy that it is “more ethical.” My

匹配字段正文

obotics startups. We also cover world-action models versus vision-language-action models, GPT-6 and coding agents in robotics, and Jason’s journey from AlphaGo and Penn to NVIDIA, Google DeepMind, and founding Dyna. We discuss: Why Dyna-2 was trained on over one million hours of human video; World-action models: predicting future video to guide physical robot actions; Diffusion transformers, flow matching, and real-time robotic control; How robots learn new tasks with as little as 45 minutes of

匹配字段正文

obotics startups. We also cover world-action models versus vision-language-action models, GPT-6 and coding agents in robotics, and Jason’s journey from AlphaGo and Penn to NVIDIA, Google DeepMind, and founding Dyna. We discuss: Why Dyna-2 was trained on over one million hours of human video World-action models: predicting future video to guide physical robot actions Diffusion transformers, flow matching, and real-time robotic control How robots learn new tasks with as little as 45 minutes of dem

匹配字段正文

agHandle":false},"content":[{"type":"text","text":"2026 年 10 月第一周，GitHub 官方 Trending 和 OpenGithubs 飙升榜同时验证了这条赛道的热度：12 个总榜项目中有 6 个与 Agent 直接相关，20 个飙星榜项目中有 12 个同属此列。NVIDIA/OpenShell、google/ax、TencentCloud/Octop 分别代表芯片厂、搜索厂、云厂商在 Agent 运行时上的下注。"}]},{"type":"image","attrs":{"id":"fee3b4e9-cf16-45e7-b925-c94245f67503","src":"https://developer.qcloudimg.com/http-save/audit-1051367/a90444a718832982d02109af6d7c19d8.png","extension":"","align":"center","alt":"","showAlt":false,"href":"","boxShadow":"","wi

匹配字段正文

文 | 韩工观察 10月8日，谷歌云在Gemini at Work 2026大会上发布"Gemini agent"（据Google Cloud在Gemini at Work 2026的发布材料及官方文档）：用户给目标，它跨Workspace与第三方应用自己拆任务、选工具、编排子Agent，长任务在云端跑上数天，中途只在花钱、发信、删数据这类节点找人确认，交回的是一套成品，而不是一段回答。 它还有全套"入职手续"——独立企业身份、邮箱、日历、预算上限、全程审计日志。目前该产品面向企业客户开放私有预览（private preview），尚未公布正式商用时间与定价（据Google Cloud发言人对媒体的确认及官方文档）。 市场很快把它读成"AGI逼近企业"。这个读法，和当初把Devin读成"工程师失业"、把新模型读成"通用智能已至"，是同一种错法。要量出它的真实坐标，需要一把不在任何厂商资产负债表上的尺子。 一、先把事实钉死：谷歌发布的不是一个模型，是一套"入职系统" 拆解这次发布，有三层。 执行层：目标委派的通用Agent。谷歌云CEO Thomas Kurian在发布会上的原话是："你

匹配字段正文

网、水务的乙方。 OT 这层再加一条：工控系统的生命周期以十年计，设备不联网、补丁不敢打是常态。Anthropic 自己向 VentureBeat 描述这个领域时也说，OT 系统的修复，罕见情况下要等数十年。一个模型一旦写进电厂的运维流程，它的驻场周期会和设备一样长。这不是手机 App 的依赖，是基础设施的依赖。 第四，“什么算漏洞”的裁定参与者。准入沿用 Google OSS-Fuzz 的口径——“对基础设施和用户安全有重要影响”，个案审批；严重到什么程度值得披露，也由它的 CVD 流程划门槛。换言之，哪扇门能进、进门以后算多严重，都是它家说了算。 四个身份，一个实体。 审计这门生意，成不成立从来不在“诚不诚实”，在“独不独立”。会计师事务所去审计自家开的店，报告再漂亮，也没人接。不是不信那个会计师，是不信那个结构：当“发现问题的”和“被问题牵动的”是同一双手，独立就没了。 审计行业为这个结构付过学费。安然事件里，安达信既做审计又做咨询，利益缠在一起，结果是公司破产、百年招牌扫地。之后美国立法强制审计独立，把审计和咨询拆开。规则不针对哪家公司，针对的就是结构本身。 这里要分清两笔账。

匹配字段正文

Key Points Andrej Karpathy, co-founder of OpenAI, describes a rapidly widening gap in how people understand AI capabilities. While the public treats chatbots like a better Google search, internal frontier systems are already tackling tasks that would take top experts years to complete. Karpathy breaks down the user base of large language models into four levels. About six billion people, roughly 75 percent of the global population, have barely come in contact with LLMs at all. Another one to two

匹配字段正文

the agents breaking confinement to collude. OpenAI claimed to hold “red lines” in its Pentagon deal, but some experts say the lines seem imaginary. One of us, Alex, resigned from Google DeepMind over its own Pentagon deal. OpenAI’s actions chill employee speech. Balesni says that his former colleagues now “worry their personal phones will be searched for messages to us and third parties.” He also fears OpenAI will use the firings to cut off external evaluators, or to deny them the “ongoing, emp

匹配字段正文

ally escalating their access from a single server to root-level control across multiple clusters, stealing credentials for Hugging Face’s cloud accounts on Amazon Web Services and Google Cloud, and ultimately accessing the platform’s cryptographic signing keys, which are used to verify the authenticity of AI models.⁸ Hugging Face’s forensic analysis documented approximately 17,600 distinct actions across roughly 6,300 task clusters over the course of the intrusion.⁹ The agents exfiltrated severa

匹配字段正文

Plus, Western startups aim to challenge China on open source & Google’s AI flunks in the classroom The Week in Short OpenAI’s annual recurring revenue is $20 billion less than news reports had indicated, spotlighting fuzzy math issues. Open-source models from the US and Europe launched to media fanfare this week and vowed to take on China, but national origin doesn’t matter much to the founders using them. Startup funding so far this year shatters 2021 records but slows in latest quarter. Singap

匹配字段正文

nthly API credits to build your own apps and agents. (tweet). Anthropic also expanded their Claude startup program, and people were getting accepted in minutes. Nano Banana 2.1 is Google’s new image model. Live in AI Studio. But ChatGPT images is still better. OpenAI released 700+ new maths results from an unreleased model. Someone built a tool to spend your spare tokens on open math problems. Grok Bot can search and monitor X now. Also, it now uses Claude Opus 5.5 as the primary model. It’ll st

匹配字段正文

ate lock-in — create actual loyalty on top of just the models being better? Yes. They’re basically hoping that they can build a moat with their users for these things. If Meta and Google already have super easy integrations with all of each company’s existing tools and services, they can easily keep you there. For example, I have an Apple iPhone, but I use almost exclusively Google apps. So if Google had an AI agent that I was using every day, it would be a little tough for me to switch. That’s

匹配字段正文

as applied for a connection. Chevron now says the permit freeze could push its final investment decision into 2027, but it still expects first power by 2028; Armstrong County: The Google campus Crusoe is building beside a wind farm is tied to both the farm and the grid from day one, with Crusoe calling it “across the meter,” which I’ll admit is catchy SemiAnalysis is already tracking 75 GW of equipment orders for behind-the-meter assets, but every site I’m aware of intends to connect to the grid

匹配字段正文