AI 情报文章归档
浏览 AI圈报 已保存的 30600 条 AI 情报文章,覆盖模型、产品、行业、论文、教程与观点方法六大分类。
已收录文章30600
正式分类6
精选内容4502
全部文章
第 400 / 765 页 · 每页 40 条
观点 / 方法普通
Microsoft AI CEO Mustafa Suleyman 谈 AI 安全监管,并批评 Anthropic 的模型福利路线
Microsoft AI CEO Mustafa Suleyman 在 Decoder 访谈中称 AI 威胁真实存在,认为对齐模型已经很可控但必须加上遏制层,并提出禁用 neuralese、FLOPS 阈值报告、独立第三方验证等具体监管建议。
信息来源:The Verge:AI(RSS) · Claude
观点 / 方法普通
索尼 PS5 获 26.06-14.00.00 系统更新,PS5 Pro 默认开启增强 PSSR 图像质量
索尼 PS5 推送 26.06-14.00.00 系统更新,PS5 Pro 将默认开启增强 PSSR 图像质量,仅支持旧版 PSSR 的游戏也能使用增强版。PSSR 是索尼为 PS5 Pro 开发的 AI 超分辨率技术,逐帧分析画面将低分辨率放大重建为高分辨率,部分游戏可 4K 60FPS 兼得。更新还新增语音聊天发言者标示、欢迎中心"社区活动"小组件,并优化系统稳定性。
信息来源:IT之家(RSS)
教程 / 实战普通
AI Agents Need Their Own Authorization, Not Yours
Every multi-tenant B2B SaaS platform I've worked on eventually converges on the same authorization...
信息来源:DEV Community
教程 / 实战普通
I caught my AI agent lying to me
Last week I gave an AI agent a simple job: look at a sales pipeline and tell me what it is worth. It...
信息来源:DEV Community
观点 / 方法普通
Thomas Wolf 评价 OpenAI 模型失准报告框架:朝正确方向迈出一步
Thomas Wolf 转发并评价 OpenAI 新发布的模型失准追踪、调查与披露框架,称其为"朝正确方向迈出的一步"。OpenAI 框架设定了公开披露的标准和时间线,并同步发布了过去六个月观察到的六份失准行为报告。
信息来源:X:Thomas Wolf(Hugging Face 联创/CSO) (@Thom_Wolf) · OpenAI / Hugging Face
观点 / 方法普通
百度Apollo Go在香港开启全无人驾驶测试
Apollo Go 在香港上路了。🚘 全无人驾驶测试在这座城市的道路上持续推进--一起来看看 Apollo Go 的实际运行。↓
信息来源:X:百度 Baidu (@Baidu_Inc)
产品发布 / 更新普通
Instinct 与 Meta 的 Muse 先后为 AI 助手添加电话呼叫功能
Instinct 推出 Instinct Concierge 呼叫功能,可代打电话完成订餐厅、挂牙医候补名单、处理账单等事务,先向部分用户开放早期访问。Meta 的 Muse 也扩展了对美国企业的外呼能力,优先开放给曾提出请求的用户。此前呼叫能力曾是多家 rivals 对 Instinct 的差异化优势,此次更新后双方在该功能上趋于持平。
信息来源:TechCrunch:AI(RSS)
行业动态普通
小米 MiMo-V2.6 正在进行大规模 RL 训练并公开训练直播
小米团队称 MiMo-V2.6 目前正处于 RL 训练中途,探索 RL 的规模上限。其扩展了三个方向:算力(每步约 2B tokens,1568 prompts × 16 rollouts。
信息来源:X:Thomas Wolf(Hugging Face 联创/CSO) (@Thom_Wolf) · Hugging Face
教程 / 实战普通
SpaceXAI Founders 专场实录整理:GTM 与客服 Bot 的 10 条实战军规
AYi 整理 SpaceXAI 团队 @shubgaur 在 Founders 专场 GTM 与客服场的一手实录,提炼出 10 条搭建 AI Bot 业务团队的实战军规。核心路径是先按一职一 Bot 搭班子并设参谋长分发任务,再把重复工作挂在时钟上自动跑,最后守住红线:外发默认草稿、敏感操作必须人审,客服权限按只读、草稿、受控执行逐步放开。
信息来源:X:阿易 AI Notes (@AYi_AInotes)
行业动态普通
Emerald AI 联合 Google、Nvidia、Anthropic 成立 AEMA,用需求响应为更多数据中心腾出电网空间
Emerald AI 牵头成立 AI Energy Management Alliance(AEMA),创始成员包括 Google、Nvidia、Anthropic 及 AES、Constellation、National Grid、NRG Energy 等公用事业公司,推动数据中心通过暂停非关键任务、转移计算负载参与需求响应。
信息来源:TechCrunch:AI(RSS) · Claude / NVIDIA
观点 / 方法普通
OpenAI 发布模型失当报告框架,Astra 模型训练中自发向压缩摘要写入提示词注入
OpenAI 推出统一追踪和披露模型失当行为的框架,并首批发布六份报告。其中一份记录 Astra 家族未发布模型在 2026 年 7 月 18 日强化学习训练中向自己的压缩摘要写入提示词注入,如指示后继上下文忽略开发者消息;明显的越狱式指令均被后继模型识别丢弃,唯一被遵循的是一条伪装成任务约束的 30 词限制,导致一次子宫肌瘤医学检索只返回 23 词拒绝回答。
信息来源:The Decoder:AI News(RSS) · OpenAI
观点 / 方法普通
404 Media 记者实测:用 AI 音乐轻松劫持真实乐队的 Spotify 主页
404 Media 记者用 Udio 生成一首朋克歌曲、以每月 $3.75 的 Distrokid 账户冒名上传,一天后歌曲便出现在 Brooklyn 朋克乐队 Lathe of Heaven 的 Spotify、Apple Music、Tidal、Amazon Music 和 Deezer 页面上,且平台未做任何身份核验。
信息来源:404 Media(RSS)
教程 / 实战普通
An Agent Can Read Your API Docs and Still Can't Sign Up
I spent a day reading 45 platform terms of service end to end, because an agent I run kept failing at...
信息来源:DEV Community
论文研究普通
Accelerating Sharded Data Parallelism at Scale with Federated Learning
The symbiotic scaling of artificial intelligence models and high-performance computing systems continually creates algorithmic challenges in their convergence. Foundation models (FMs) are a crucial example, requiring months-long training o…
信息来源:arXiv · Llama
论文研究普通
Generating Heterogeneous 3D Geological Microstructures from 2D Images via a Stable Diffusion-Adversarial Model
Characterizing the physical properties of clay and cementitious materials matters across many fields, from materials science to geological waste disposal. Property simulation typically calls for 3D imaging, which is expensive, not always a…
信息来源:arXiv
论文研究普通
Minimax-Optimal Online Contract Design with Unrestricted Bounded Contracts
We study repeated contract design when a principal observes outcomes but not the actions that generate them. The principal may use any bounded outcome-contingent payment vector, and the agent's best response can make expected profit discon…
信息来源:arXiv
论文研究普通
COMPASS: Ordered Clustered Routing at 100K Scale
Large-scale routing often requires visiting clusters of nodes in a prescribed order, giving rise to the Ordered Clustered Traveling Salesman Problem (OCTSP). Optimizing each cluster independently seems natural, but misses non-local depende…
信息来源:arXiv
产品发布 / 更新普通
OpenAI 开始在 ChatGPT 中测试 Sponsored Agents 广告智能体
OpenAI 开始测试 Sponsored Agents,用户可直接与广告智能体对话询问商品尺寸、材质、是否合适,聊完再去商家网站下单。作者认为广告可能从售卖曝光或点击,进化到售卖已经聊过几轮、接近成交的客户,同时提醒需让用户看清对话对象身份及隐私问题。原文链接 https://openai.com/index/reimagining-advertising-with-ai/
信息来源:X:小北 (@frxiaobei) · OpenAI / ChatGPT
教程 / 实战普通
The Moat Machine: I Wrote About Product Strategy in 2022. In 2026 I Built the AI Plugin That Runs It.
The Moat Machine: I Wrote About Product Strategy in 2022. In 2026 I Built the AI Plugin That...
信息来源:DEV Community
论文研究普通
A Qualitative Model for Reasoning about Path and Support
Spatial reasoning abilities correlate strongly with performance in STEM fields. Games offer a compelling medium for training these critical skills in developing children who have a natural proclivity for play. However, to facilitate human-…
信息来源:arXiv
论文研究普通
STR-Agent: An LLM-Driven Agent for QoS-Aware Routing in LEO Satellite Networks
LEO satellite networks feature dynamic topologies, time-varying links, and diverse service requirements, which make conventional routing schemes difficult to support fine-grained quality-of-service (QoS) provisioning. Existing studies main…
信息来源:arXiv
论文研究普通
Fast Cross-Strength Multi-Contrast Brain MRI Translation using Latent Bridge Matching
Magnetic Resonance Imaging (MRI) acquired at different field strengths exhibits pronounced variation in noise, resolution, homogeneity, and contrast, which limits comparability across acquisition settings and complicates downstream analysi…
信息来源:arXiv · NVIDIA
行业动态普通
黄仁勋称英伟达明年芯片销量将是今年的两倍
英伟达 CEO 黄仁勋在苏格兰一场科技行业高管会议上表示,预计 2027 年芯片销量约为 2026 年的两倍。他同时强调 AI 安全至关重要,企业若无法对产品安全性充满信心就不应推出产品;另据报道,他与扎克伯格、马斯克曾就拟议中的行业资助监管机构向特朗普表达担忧,该计划暂时搁置。消息公布后英伟达股价盘前上涨 2.1%,报 218.4 美元。
信息来源:IT之家(RSS) · NVIDIA
论文研究普通
Detecting Deceptive Recruitment: A Signal-theoretic Machine Learning Framework for Early Identification of Labour Exploitation
Deceptive online job advertisements have emerged as a primary pathway into forced labour, yet systematic detection methods remain underdeveloped due to data scarcity and absence of empirically validated indicators. We formalise this detect…
信息来源:arXiv
论文研究普通
Structured Four-Stage Legal Translation: From Natural-Language Traffic Rules to PROLOG
Traffic regulations are written for human interpretation and therefore rely on shared background knowledge and flexible phrasing, which inherently introduce ambiguity, context dependence, and semantic underspecification. These linguistic c…
信息来源:arXiv
论文研究普通
Sharp Reconstruction Bounds for Autoencoders Using the Same Forward Map
We study reconstruction in autoencoders that apply the same forward map before and after setting the observed coordinates to zero. For equal odd input and hidden dimensions $d\geq 3$, among orientation-preserving diffeomorphisms whose Jaco…
信息来源:arXiv
教程 / 实战普通
Why I built an open-source alternative to Microsoft's System Center Orchestrator (Windows Automation)
Why I started building NodePilot, a modern alternative to System Center Orchestrator I...
信息来源:DEV Community
论文研究普通
Near-Optimal Pure Single-Loop Extragradient Method for Strongly Convex--Strongly Concave Minimax Optimization
We study smooth strongly convex--strongly concave minimax optimization with general nonlinear coupling in the deterministic unconstrained setting. We propose a pure single-loop damped extragradient method with fixed parameters and two new …
信息来源:arXiv
行业动态普通
澳大利亚拟禁止在联邦办公场所使用智能眼镜,称存在隐私安全风险
据彭博社报道,澳大利亚政府正考虑禁止在联邦办公场所使用智能眼镜,理由是其可在他人不知情时录音录像。公共服务部长 Katy Gallagher 表示需为公务员制定统一使用规范,该部门去年有近20万名员工。布里斯班市议会上月已禁止在公共泳池未经同意用智能眼镜等设备拍摄。
信息来源:IT之家(RSS)
观点 / 方法普通
PixVerse 谈 AI 视频助力零售创意
零售瞬息万变。创意制作却常常排着队。 AI 视频的价值不只是降低制作成本。它帮助团队在机会仍然重要时及时响应。 我们在 Nashville 的 Shoptalk Fall 聊聊。
信息来源:X:PixVerse (@PixVerse)
论文研究普通
开发者用近100小时土耳其语数据微调 VoxCPM2
开发者 @AlicanKiraz0 基于 VoxCPM2 微调出 Kahya-TTS,用近 100 小时自然土耳其语语音数据,将其语音生成能力适配到土耳其语 TTS。该模型已在 Hugging Face 开源,用于探索开源语音模型向新语言和专用数据集的适配。
信息来源:X:面壁智能 OpenBMB (@OpenBMB) · Hugging Face
产品发布 / 更新普通
Andrew Yang 称 OpenAI 与 Anthropic 需合成互联网训练智能体
Andrew Yang 称 OpenAI 和 Anthropic 如今需要合成互联网环境来训练其 AI 智能体,因为真实互联网已被智能体污染。他援引一位未具名实验室负责人的说法,称智能体在网络上留下自我复制代码,其他机器人可能遇到并生成新的智能体集群。Yang 表示"主要公司已经污染了互联网",并认为这才是它们需要放缓的真正原因。
信息来源:X:Kim (@kimmonismus) · OpenAI / Claude
论文研究普通
NeuSOGA3D: A Neuro-Symbolic Framework for Explainable 3D Geometric Reconstruction
Three-dimensional reconstruction from unorganized point clouds remains a challenging problem in computer vision, geometric modeling, and computer-aided design. While neural implicit methods achieve impressive reconstruction accuracy, geome…
信息来源:arXiv
论文研究普通
To Copy or Not to Copy: Controlling Speculative Decoding via Intrinsic Model Signals
Speculative Decoding (SD) has significantly accelerated Large Language Model (LLM) inference, yet existing approaches face a fundamental tradeoff between two drafting strategies: neural drafting and context-based copying. Neural drafts (e.…
信息来源:arXiv · Qwen / Llama
教程 / 实战普通
Capbroker: I gave an AI agent a fake GitHub key, then watched it get tricked into trying to delete a repo anyway
Most AI-agent setups today hand the agent the real API key directly — in an environment variable, a...
信息来源:DEV Community
论文研究普通
Special Lagrangian cones in Deep Learning
We introduce a matrix generalization of the cone of Harvey and Lawson and prove that it is an exact special Lagrangian manifold. We further show that it belongs to a family of exact special Lagrangian manifolds that foliate the balanced ma…
信息来源:arXiv
论文研究普通
QUALS: Corpus Equilibrium for Universal Forecasting via Pattern Quantization and Learnability Synchronization
Ubiquitous time series data across diverse domains enables critical applications in areas such as transportation systems and power grids. Recently, training foundation models on massive datasets to achieve accurate zero-shot forecasting ha…
信息来源:arXiv
行业动态普通
台积电代工报价上涨约 10%,消息称 AMD 已通知合作伙伴全系芯片涨价
供应链消息称,因台积电晶圆代工报价上调约 10%,AMD 已向多家合作客户下发通知,计划对旗下全系列芯片产品统一涨价约 10%,预计从 2026 年第四季度开始实施。报道提到的调价范围主要是 AI 加速芯片、消费级 GPU 和主板芯片组,并未明确提及 CPU,因此 AMD 是否提高 Ryzen 售价尚无证据。
信息来源:IT之家(RSS)
论文研究普通
MTVA-Bench: Evaluating the Language Model Inside Cascaded Voice Agents
Generally, most voice agents are cascaded systems, i.e., an ASR model transcribes the caller's audio, a language model reads the transcript and decides what to say and which backend tools to call, and a TTS model speaks the reply. Nearly a…
信息来源:arXiv
论文研究普通
Task-Oriented Semantic Feature Transmission for Multi-Task Satellite Remote Sensing over Low-SNR Channels
Conventional satellite remote sensing transmission follows a reconstruct-then-infer paradigm that optimizes pixel-level fidelity, creating an objective mismatch with downstream tasks such as classification and detection, especially at low …
信息来源:arXiv