跳到主要内容
每日新闻流·2026-11-09 — 2026-11-10·中文 / 英文原文

AI每日新闻流 · 2026-11-10

本期资讯逐条整理,保留可核对的原始来源。

60 条事件64 条来源记录

AI 模型

13

基于 Kubernetes 的 LLM 分布式推理与 llm-d、Wasm 及 TornadoVM 集成——InfoQ

资讯发布:

在本期节目中,Edoardo Vacchi 探讨了 WebAssembly 在后端应用的演变趋势,包括利用 TornadoVM 在 GPU 上运行 Java 工作负载,以及使用 llm-d 和 vLLM 在 Kubernetes 上优化分布式 AI 推理。Edoardo 解释了 llm-d 如何管理缓存局部性(KV 缓存和前缀缓存)、请求路由,并在 Kubernetes 集群中实现跨节点的批量推理,以最大化昂贵的 GPU 利用率。

Amazon Bedrock 托管知识库支持单次检索与智能体检索 | 人工智能

资讯发布:

Amazon Bedrock 托管知识库现在支持通过 `AgenticRetrieveStream` 进行单轮检索增强生成(RAG)和仅检索请求。在单轮 RAG 模式下,服务会一次性搜索您的知识库,并使用基础模型(FM)根据找到的段落生成答案。

Pinterest 用 40 亿 Qwen3 LLM 取代 Journey 聚类,报告参与度提升 - InfoQ

资讯发布:

Pinterest 将其用户旅程推理系统从多阶段聚类管道迁移至经过微调的 40 亿参数语言模型,报告显示邮件点击通过率提升 1.1%,推送打开率提高 1.3%。

Anthropic 发布新 Mythos 和 Fable 模型,引发 RSI 担忧 | TechCrunch

资讯发布:

Anthropic 于周一发布了其最先进模型的新版本,包括 Mythos 5.2 和 Fable 5.2。公司表示,新模型在编程、推理和视觉等方面“显示出改进”,并能更清晰地与用户沟通。

Google Releases Gemini 4 Argon for Robotics Control, Simulation, and Policy Improvement - InfoQ

资讯发布:

Google has released Gemini 4 Argon framing it as a general-purpose physical-reasoning model for robot control, Real2Sim scene construction, multimodal labeling of robot video, and iterative policy improvement, rather than as a single end-effector policy.

Taiwan-Based Hacker Used BOSSAI AI Agents to Steal Taipei Emergency Dispatch Records

资讯发布:

A single attacker using AI agents pulled emergency dispatch records covering seven Taipei districts, researchers at Acronis's Threat Research Unit said last week.

Frontier AI Cyber Risks: Bank of England Warns on Impact

资讯发布:

Bank of England Governor Andrew Bailey is warning that cyber risks from frontier AI extend beyond faster attacks: the technology can also make them cheaper, expand their reach and put sophisticated capabilities in more hands. He is calling for tougher testing of AI models before deployment and after release, alongside a societal right to intervene in deployment decisions.

What Hardware Does Your PC Need To Run A Local AI Model?

资讯发布:

Less than you might think, but a 16GB PC will only take you so far.

Deconstructing the AI Trade: AI Builders, AI Supplies and AI Products!

资讯发布:

Common and Competing Interests!

Lawyer builds private AI server with eight used Nvidia V100 GPUs — up to 215 tokens per second to keep legal files offline, costing about $2,400 per four-GPU board | Tom's Hardware

资讯发布:

A lawyer posted on Reddit about their private AI server built from eight used 2018-era Nvidia Tesla V100 data center GPUs so case files never have to leave the office. With two 4-card SXM2 NVLink boards, four PEX switches, and SlimSAS cables everywhere — plus one burnt power connector — the builder now runs the open DeepSeek V4.1 Flash model with every expert in GPU memory, at up to 215 tokens per second.

Wearable sleep trackers are using your brainwaves to train AI models without explicit user consent — tens of millions of hours of EEG data harvested | Tom's Hardware

资讯发布:

Science recently conducted an investigation alongside Pulitzer Center's AI Accountability Network and found that the customer neurology companies have accumulated tens of millions of hours of electroencephalography (EEG) recordings — a rather technical way of saying brain-wave data. As per the findings, several are using those recordings to develop so-called brain foundation models, despite initially failing to spell out that use in their privacy policies.

Disney's AI deal with Adobe is raising eyebrows | Creative Bloq

资讯发布:

Just as Adobe MAX 2026 gets underway in Miami, the maker of Photoshop and Illustrator has announced an eye-catching deal that could change how the world sees it. It's now an Official AI Innovation Partner of The Walt Disney Company.

[$] Git considers voting to determine its LLM policy

资讯发布:

One of the dominant themes of 2026 has been the debate within development projects about the terms under which LLM-written contributions should be accepted — if, indeed, they should be accepted at all. The Git project has not been immune to this trend; the topic was discussed at the recent Git Contributor's Summit.