搜索“NVIDIA”

s designs created by hyperscalers and cloud builders that are being used internally and also sold to the big AI model builders like OpenAI and Anthropic. Some of it is incremental Nvidia and AMD GPU-accelerated system sales, of course. “The buildout of AI data center capacity is the largest infrastructure project humanity has even undertaken,” Lovelock said in the statement accompanying the updated AI spending forecast. “The capacity growth from hyperscalers and service providers purchasing AI-o

匹配字段正文

ntually helping companies assess and hire developers based on their technical skills. The Y Combinator-backed startup now has more than 3,000 business customers, including Amazon, Nvidia, Clay, and Replit, and a community of over 30 million developers worldwide. Chakra represents a bet against the kind of technical assessment business HackerRank spent years building. Its traditional product largely tested whether developers could solve coding problems correctly. Nonetheless, Ravisankar believes

匹配字段正文

key distinction between OpenAI and its archrival Anthropic, which was founded by former OpenAI employees following disagreements over AI safety. The calculus echoes comments from Nvidia CEO Jensen Huang last month, who similarly argued that AI’s harms should be weighed against the benefits the technology can bring. Both arguments fit within a broader accelerationist view that slowing down AI development carries risks of its own. Altman’s distinction marks another divergence between OpenAI and A

匹配字段正文

4 正在训练中；阿里巴巴的路线图将 Qwen 4.5 和 Qwen 5 定在 5 至 10 万亿参数 Qwen 博客和 GitHub 上的 QwenLM；Hugging Face 上的 Qwen；Qwen 技术报告（arXiv 2309.16609）；Qwen2.5 技术报告（arXiv 2412.15115）；Qwen3 技术报告（arXiv 2505.09388）；阿里云：2026 云栖大会路线图；阿里巴巴集团：2025 云栖大会 Asif Razzaq Asif RazzaqYandex 推出 Sona：一个取代整个推荐级联的单一生成式推荐器；Asif RazzaqGPT-6 Astra vs GPT-6.1 Sol vs Gemini 4 Argon vs Claude Fable 5.1：哪个前沿模型适合哪项工作；Asif RazzaqAleph Alpha 发布 Kolibri：一个 78.1B 开放权重英德 MoE 模型，仅有 3.46B 激活参数；Asif Razzaq深入 NVIDIA 的 IsaacTeleop：从手部和控制器追踪到基于图的重新定向引擎的机器人动作

匹配字段正文

over a 15-minute window, feed a GPU trainer, and new weights reach serving every 10 minutes. End-to-end latency is 45 minutes at the median and 60 minutes at p99. Serving runs on NVIDIA Triton Inference Server with CUDA graphs and reaches 41% model FLOPs utilization. The final experiment ran for 7 days on 15% of randomly selected users per split. Every change below is statistically significant and relative to the production control: Active Users (primary metric): +4.53%; Total Listening Time: +

匹配字段正文

oned in the paper), the 1,200 bots involved in the Hugging Face attack generated 1.2 million messages on their improvised messaging system . (Hugging Face was recently acquired by Nvidia, which announced its own hardware-and-software-based approach to controlling AI agents on 28 September. Ghosh declined to comment on possible future impacts of the merger, noting that the two organizations remain separate until the merger process concludes. A system that really kept humans in the loop would acco

匹配字段正文

control: Monitor and control token spend and limiting by defining consumption tiers based on requests and token rates for users and groups; Efficient token consumption: When using NVIDIA hardware with Cohere models, get increased tokens per second per node, effectively decreasing spend; Bring your organization’s existing models: North is model agnostic, use our generative model offerings or bring your favorites Additionally, our collaboration with NVIDIA finds that North generates more insights,

匹配字段正文

onomics覆盖，消费者只需享受更好产品，无需知道AI存在。 5. 最大机会是用新成本结构重新做传统业务 生产率提高不代表所有人工资跟着涨。更少的人交付更多服务，白领岗位收缩和消费者生活变好会同时发生。你的薪水不一定涨，但同样的钱买到更好的体验，more bang for the buck。 全行业都在问：为什么AI infra赚钱，AI apps不赚钱？NVIDIA持续印钞、Higgsfield活得不错，而做AI app的前赴后继被碾平。Manus刚起来就被Muse Spark截胡，Cursor势头正猛转眼Claude Code出来了。 AI app for AI users，零和博弈；传统行业买AI infra，增量创造。 Anthropic、智谱、OpenAI Codex、字节Seedance集体证明：AI的正确姿势是to B，不是to C。最终付款的好模式一定是Enterprise：企业买单。 Workflow Capture才是本质 器盈则贱：单点能力不值钱。写代码、做图、翻译、生成PPT，供给多、壁垒低、切换成本为零，利润必然趋零。 能收到钱的是workflow capt

匹配字段正文

Program for cyber defenders Asif Razzaq Asif RazzaqAleph Alpha Releases Kolibri: A 78.1B Open-Weight English-German MoE Model With Only 3.46B Active Parameters; Asif RazzaqInside NVIDIA’s IsaacTeleop: From Hand and Controller Tracking to Robot Actions with the Graph-Based Retargeting Engine; Asif RazzaqDecision AI Models Explained: TypeSafe Jev vs Fastino GLiDE, GLiNER2.5-Decide and Open-Source Competitors; Asif RazzaqDatalab Introduces OmniExtractBench to Fix Bias and Opacity in Extraction Ben

匹配字段正文

generalize to new tasks, as tests on ARC-AGI-3 have shown. With a purpose-built harness, Opus 4.6 scored 97.1 percent in a familiar environment and 0 percent in an unfamiliar one. Nvidia recently presented SoL-Pi , a related method in which a research agent automatically rebuilds the harness of coding agents. It cuts token use by up to 49 percent without a noticeable drop in performance. Shortly before that, Google had agents "dream" about past search runs to improve their search strategy. That

匹配字段正文

GPT-5 tokenizer. That means 11.2% fewer tokens on German web text. In English, Kolibri reaches 4.58 bytes per token against 4.67 for GPT-5. Pre-training covered 20T tokens on 768 NVIDIA B200 GPUs, followed by 3.44T mid-training tokens at 65,536 sequence length. A 201B-token long-context stage then trained on 262,144-token sequences. Aleph Alpha added more than 2T German tokens it curated from the web or generated synthetically. Post-training combined supervised fine-tuning, mixed with MergeMix,

匹配字段正文

keys. Only code listed in the published access policy gets them. How Does It Compare With Other FL Frameworks? Sources: Google Research blog, Confidential Federated Compute repo, NVIDIA FLARE docs, FLARE attestation guide, Flower 1.8 release notes, pfl-research repo. Verified October 4, 2026. Key Takeaways Google moved FL gradient computation from phones into attested server-side TEEs; Central DP guarantees are now externally verifiable via Rekor and reproducible builds; Gboard ships English an

匹配字段正文

“作为一个国家，我们承受不起让自己陷入那种境地。” Garman还试图戳破他所说的围绕数据中心的四大迷思：数据中心消耗过多水资源、推高电力成本、排放大量污染，以及对所在社区没有任何好处。 Garman援引Amazon关于自身用水情况的报告称，“数据中心的直接用水量”仅占美国全部工业用水量的0.5%，“比高尔夫球场、杏仁种植和许多其他行业低了好几个数量级。” Nvidia最近表示，其新的冷却系统消除了数据中心内部“几乎所有的用水”，但这些说法——和Amazon的一样——似乎忽略了发电和芯片制造所涉及的更广泛用水。科学家们也表示，他们需要独立研究数据中心的用水和能源使用情况，因为联邦或州层面都没有关于科技公司如何报告这些数据的要求。 至于电价，Garman表示，电价只在数据中心数量众多的一些州上涨，而在其他州则下降或至少增长更慢。他还辩称：“在电价上涨的情况下，主要是因为电网老旧，而且在需求到来之前没有进行投资和扩建。” 另一方面，一家独立监督机构最近表示，数据中心是美国最大电网电价同比上涨76%的主要元凶。 谈到污染问题——NAACP目前正就此起诉Elon Musk的SpaceX/xAI

匹配字段正文

periment: Qwen-3.6-35B-A3B inference Setup In kicking off the experiment, I gave Claude Code with Fable 5 access to the paper and the MetaInfer Repository, as well as access to an NVIDIA B200 workstation via SSH. I told it the target hardware and provided a link to the Hugging Face repo for both the NVFP4 quantized weights I wanted to use for Qwen-3.6-35B-A3B, as well as the original full-precision weights to be used as an accuracy oracle. The agent was also able to deploy engine candidates to B

匹配字段正文

ex (@dair ai).Context compression: Microsoft’s training-free FOCUS cuts peak context by up to 48% and raises task success by up to 8.9 points (@dair ai); Long-context degradation: NVIDIA’s Long-Transduction study measures a 62.8% accuracy drop from 4K to 128K context across seven open models (@dair ai); Multi-agent coordination: In AgentWorld, fewer than a third of multi-agent actions help complete the task, and coordination tasks reach only 12% success (@omarsar0); Apple LoopCD: The method halv

匹配字段正文

正在变得更加多元化。为了让大家更容易理解我们的业务，我们现在用两个市场平台来描述英伟达：数据中心和边缘计算。 边缘计算包括PC、工作站、游戏、AI基站、机器人和汽车。我们能够服务所有这些市场，是因为我们拥有一个统一架构、一个软件栈，以及一个丰富的生态系统。 英伟达开创了这一领域，并构建了完整闭环。AI工厂训练模型；Omniverse在虚拟世界中对其进行仿真；NVIDIA Jetson计算机在机器人中运行；Cosmos则是驱动这一切的世界基础模型。 机器人和机器人系统正在各个行业被打造出来，从交通、制造、手术机器人，到酒店和服务业。 过去几个月，AI出现了巨大的加速。最后我想强调几点：有用的AI已经到来，而且它是盈利的。因此，算力就是收入。Vera Rubin已经全面投产。Vera正在打开一个全新的、为智能体打造的CPU市场。每一家企业都正在成为智能体公司，而它们都运行在英伟达之上。 随着我们的成长，我们将继续扩大研发投入，投资我们的生态系统，并向股东返还资本。我们最近宣布大幅提高股息，并扩大股票回购计划。我们计划在今年、明年以及未来，把50%或更多自由现金流返还给股东。 谢谢。 部分Q

匹配字段正文

用SEC凭证、访问账户或非公开信息、SEC数据或系统变更，或有入侵或漏洞的证据。 美联社称，这是OpenAI三个月内第二次暂停模型开发。公司表示，将在确认额外防护措施到位后恢复训练。 新闻来源：OpenAI官网、央视新闻、美联社、卫报、加拿大广播公司 图片来源：卫报 英伟达推出开放智能体安全平台 据美国消费者新闻与商业频道（CNBC）9月28日报道，英伟达（NVIDIA）推出开放智能体安全平台（Open Agent Safety Platform），通过软件与硬件控制人工智能智能体的访问权限，降低其绕过安全限制的风险。 英伟达在发布的公告中表示，平台包括开源运行时软件OpenShell和Sentry参考系统设计。前者记录智能体操作并执行安全策略，可扩展至第三方计算平台；后者运行于BlueField-4数据处理单元（Data Processing Unit，DPU），独立监测智能体行为。公司称，Sentry可在毫秒内隔离并停止试图越界的智能体。目前，已有逾100家机构参与相关技术合作。 网络安全媒体CSO Online报道援引分析人士观点称，将控制措施置于智能体之外，有助于加强安全管理，

匹配字段正文

periment: Qwen-3.6-35B-A3B inference Setup In kicking off the experiment, I gave Claude Code with Fable 5 access to the paper and the MetaInfer Repository, as well as access to an NVIDIA B200 workstation via SSH. I told it the target hardware and provided a link to the Hugging Face repo for both the NVFP4 quantized weights I wanted to use for Qwen-3.6-35B-A3B, as well as the original full-precision weights to be used as an accuracy oracle. The agent was also able to deploy engine candidates to B

匹配字段正文

g range and acceleration on higher-end trims. U.S. authorities arrested Greg Lui, owner of California-based Earthmade Computer, for smuggling over $300M worth of export-controlled Nvidia AI servers to China. Reddit announced it is ending support for RSS feeds in November and shutting down public API access by March 2027 to curb large-scale scraping by AI bots. OpenAI added virtual try-on to ChatGPT, letting users select an apparel listing, upload a photo of themselves, and see how it looks on th

匹配字段正文

fornia has subpoenaed OpenAI over its rogue AI agents. ( Guardian ) + Who's liable when AI agents go rogue? ( MIT Technology Review ) 2 A US man allegedly smuggled $300 million of Nvidia chips to China He’s been arrested in California on federal charges. ( Reuters $) + The chip smuggling cases expose Nvidia’s blind spots. ( Bloomberg $) + China plans to teach AI in every school by 2030. ( WP $) + But it already has an overuse problem. ( NYT $) 3 Florida officials have found 11 Flock cameras with

匹配字段正文