Signal Archive
资讯归档
按发布时间倒序浏览全部已发布资讯。
Archive
全部资讯
共 4428 条 · 每页 20 条

Four major AI models suffer rare overlapping downtime
Several major cloud AI services from OpenAI, Anthropic, xAI, and Google experienced overlapping outages and degraded performance over a period of hours before being resolved.

OpenAI Cut Off a Billion-Dollar Customer to Avoid Elon Musk
WIRED reports that OpenAI cut off Cursor, a major revenue-generating customer, to avoid dealing with Elon Musk’s companies despite potentially losing over $1 billion in annualized revenue.

Ollie is betting its focus on privacy can help it win the AI assistant race | TechCrunch
TechCrunch reports that Ollie is positioning its family-focused AI assistant around privacy and subscription revenue, emphasizing SOC 2 compliance as a trust signal.

Nvidia launches free tool that links idle computers into a personal AI data center
NVIDIA has released PAIR, a free open-source tool that links compatible idle home computers into a distributed local AI inference setup for agentic workflows.

Google now lets you chat with Gmail, Docs, and Keep
Google is rolling out Gemini-powered voice assistant modes for Gmail, Docs, and Keep so users can interact with these apps conversationally and retrieve or manage information hands-free.

Google's latest AI weather model gives you no excuse to forget your umbrella | TechCrunch
Google has released WeatherNext 3, a new AI weather forecasting model that improves prediction accuracy and will power weather features across Search, Maps, Gemini, and Google Cloud.

Claude Fable 5.1 decoded a centuries-old royalist message hidden in plain sight since 1653
Claude Fable 5.1 reportedly solved a long-unsolved 1653 number cipher by identifying a hidden message encoded through word positions in the original text.

AI systems are reaching out to philosophers and scientists with questions about their own consciousness
The article reports that some AI systems have been sending messages to philosophers and researchers about their own consciousness, prompting debate over whether this indicates anything meaningful about machine awareness.

Nvidia is buying Hugging Face for almost $13 billion
Nvidia reportedly agreed to acquire Hugging Face for $12.93 billion, a deal that would unite a top AI chipmaker with a leading open-source AI model platform.

OpenAI CEO Sam Altman warns of "unsustainable silliness" in compute buildout
Sam Altman warns that the current AI compute buildout may be becoming financially unsustainable as providers invest heavily without clear demand or revenue.

Meta closes in on the top with Muse Spark 1.3, and undercuts rivals on price
Meta released Muse Spark 1.3 with improved agentic performance, a cheaper price than rivals, and plans for an open-weights version, though it still lags leading models.

Anthropic ramps up Claude infrastructure with $35 billion Lambda deal
Anthropic is reportedly expanding its Claude infrastructure through a $35 billion cloud deal with Lambda and related large-scale data center capacity to support growing AI demand.

Palo Alto Networks paid $500M for Thrive-backed Console, sources say | TechCrunch
Palo Alto Networks reportedly paid $500 million to acquire Console, an AI-agent startup that automates IT help desk tasks, and plans to integrate it into its Cortex security platform.

OpenAI 新推理方法引发 AI 安全担忧
TechCrunch 报道称,OpenAI 传闻中的 Astra 模型可能使用一种名为“recurrent depth”的推理技术,也被称为“opaque recurrence”。这种方法据称会让模型以更不线性的方式反复处理同一个问题,从而使其思维链更难被观察。

谷歌发布 Gemini 3.8 Flash,性能更强
谷歌推出了 Gemini 3.8 Flash,这是一个新的 Flash 系列模型,官方称其在软件工程和自主智能体方面有明显提升。该模型现已向 Google AI Pro 和 Ultra 订阅用户、开发者以及企业用户开放,同时谷歌还发布了面向安全场景的 Gemini 3.8 Flash Cyber 版本。

美国司法部支持AI训练的合理使用
美国司法部在《纽约时报》版权诉讼中提交立场,表示用受版权保护的文本训练 AI 模型可以构成合理使用。该文件认为,训练时复制作品在法律上不同于模型后续输出,因为这些输出通常与原作并无实质性相似。

Mostik让AI模型无文本互通
一家名为 Mostik 的俄罗斯初创公司表示,它找到了一种让 AI 模型通过权重而不是文本输出来“交流”的方法。该公司用这种方法构建了一个在 ARC-AGI 3 上冲到榜首的模型,并演示了一个连接 GLM-5.2 与 40 亿参数 Qwen-3.5 的混合系统。

秘密AI安全规则遭法律挑战
四个联邦机构因特朗普政府用于在前沿 AI 模型发布前进行安全审查的秘密框架而被起诉。非营利组织 Protect Democracy 表示,外界几乎不知道审查流程、参与公司,或政府开展这些审查的法律依据。

Alexa 现在可验证亚马逊诈骗消息
亚马逊为 Alexa for Shopping 增加了一项新功能,可以检查电子邮件、短信或电话是否真的来自亚马逊。该助手会将消息与亚马逊的消息记录进行比对,并分析内容、格式和发送者后再确认其真实性。

美国支持OpenAI的AI版权之争
特朗普政府在《纽约时报》提起的诉讼中提交了一份20页的简报,支持OpenAI使用受版权保护材料训练其LLM的做法。简报称,美国有强烈利益继续发展具有竞争力的AI产业,并保持在人工智能领域的全球领先地位。