AI 新闻 · V1

关于 AI 灭绝风险几乎没有严肃研究。为什么？也许这可以成为第一步？

Astra 做的梗图已经比 redditors 还好了，完了。

任何认真对待「AI 即将导致人类灭绝」说法的人，都应该读一读这个帖子串。也请关注此人；他一直是 AI 领域最犀利的评论者之一。

很高兴分享我们在 @GoogleAIStudio 中面向人类与智能体的完全集成文档体验！！这件事我想了两年半，抱歉花了这么久，但这是迈向进一步重塑体验的第一步。感谢 @timeyoutakeit 和团队的出色工作。

你现在在数学领域看到的情况，是「锯齿状前沿」的结果，也是其他职业即将面临的预兆。是的，数学家做数学，但他们还有其他自认为重要的工作（指导学生、维护科学共同体、守护学科未来、培养对数学的热爱），而这些是 AI 做不到的。至少有一种担忧是：AI 公司聚焦于数学家工作中最显眼的部分（写证明），正在损害那些 AI 做不到的其他任务。AI 可以发现超人类的证明，但那并非数学职业的全部，实际上可能削弱并减少外界对其他重要工作的关注。如果最显眼的部分被拿走，就更难向外界捍卫数学家其他任务的价值。我怀疑我们会在越来越多的领域和职业中看到这种情况：人们将被迫帮助他人理解，他们的工作不仅包括 AI 能做的那些最显眼任务，还包括 AI 做不到或做得很差的任务。

面向临床医生的 GPT-6 Astra Pro：（引用称经认证的美国临床医生今日可通过 ChatGPT for Clinicians 免费获得 GPT-6 Astra Pro 访问权限）

Grok @Bot（引用 Grok Bot 关于近期体验改进的帖子：可让 Bot 先起草消息，经你批准后再发送）

现已推出：面向金融服务的 ChatGPT。这是一款定制的 ChatGPT Work 体验，将内置金融数据与 GPT-6 Astra 的推理能力相结合。团队可用于开展研究、构建财务模型，并制作定制化客户材料。

我们发布了迄今为止最详尽的威胁情报报告。报告涵盖了人们试图滥用 Claude 的方式——网络攻击、影响力行动、监控、生物以及制造武器——以及我们如何发现并阻止了这些行为。我们已瓦解报告中的所有行动，并用这些经验强化了防护措施。在适当情况下，我们也将发现通报给当局及其他 AI 公司。这些案例并不典型：我们突出的是所见过的一些最复杂的滥用。但它们尤其值得讨论，因为它们揭示了 AI 滥用的走向、我们的防护在何处有效、以及需要改进之处。发布此报告是为了让其他人能在自有平台上发现同类活动，也让公众更清楚地了解新兴威胁如何演变。

用 Search 为下一场重要比赛做准备的 3 种方法

现在人人都能让数据发挥作用

一位研究者如何用 Codex 和 ChatGPT 搜索新型抗菌分子

Simultaneous assessment of medical imaging and patient records is often required in clinical diagnosis. However, standard machine learning algorithms cannot analyze these data types together. Meanwhile, compliance with HIPAA and GDPR can constrain centralized aggregation of sensitive patient data. This leaves a crucial void of secure fusion of visual and textual context across distant networks. Th

We present the PACE, a framework for retrieval-augmented dialogue serving that formalizes Perceived Time-to-First-Response (PTFR) as a QoE objective and minimizes it under quality/cost constraints. Unlike prior work on cascaded routing, semantic caching, or adaptive retrieval, PACE jointly controls which answer source composes the response and what fills the waiting window. Deployed on a humanoid-

Finding new physics (NP) is the most important problem in particle physics today. Studying ``anomalies'', i.e., measurements of low-energy observables whose values disagree with the predictions of the Standard Model (SM), is a powerful search strategy. The SM Effective Field Theory (SMEFT) provides a general model-independent framework for parameterizing NP; it is natural to try to find the SMEFT

Animation and VFX pre-production review requires teams to translate loosely specified creative intent--briefs, evolving specifications, heterogeneous references, and verbal decisions--into revisions that junior artists can execute without repeated clarification. In practice, criteria drift across iterations, review judgments lose their evidential basis, and the reasoning behind a request rarely su

This paper describes the Rosetta system for Subtask 1 (Context-Aware English-to-Dialectal Arabic Dialogue Translation) of the AlexandriaX shared task, participating in both constrained and unconstrained tracks. The approach fine-tunes a LoRA adapter on NileChat-3B using structured system/user prompts that condition generation on dialect and dialogue context. For the unconstrained track, the adapte

Exception Related Code (ERC), which includes throw statements, conditions (if statements) that guard those throw statements, and try/catch blocks, is an essential component of software systems, allowing developers to detect and handle exceptional states that deviate from the expected program behavior. However, manually writing ERC across large codebases is tedious. We propose a novel task: retrofi

Cross-silo Federated Learning (FL) enables geographically distributed institutions to collaboratively train machine learning models without sharing raw data. In wide-area deployments, however, communication delays often dominate round completion time and exacerbate the straggler effect. Hybrid FL addresses this challenge by combining synchronous and asynchronous client participation, but effective

The growing complexity of content moderation policies presents a critical challenge for their consistent operationalization. While foundation models possess the basic capabilities needed to confront this challenge, whether they can reliably moderate online content remains an unanswered question. In this paper, we systematically compare two competing paradigms for Vision-Language Model (VLM) guidan

Current LLM memory systems treat all personal facts identically, so stores grow without bound while retrieval precision degrades. The core challenge is lifecycle management: which memories should persist, which should be replaced, and at what rate, conditioned on the behavioral type of each fact. Fortunate Recall (FR) is a composable policy layer that classifies personal facts into a 10+1 behavior

Background: Emergency Department (ED) return visits are commonly reviewed for quality assurance, but are often limited (e.g., to revisits within 48-72 hours) to increase actionable finding yield while minimizing chart review burden. Those limitations may lead to missed quality improvement opportunities. Methods: We conducted an exploratory, retrospective study of randomly selected ED visits to a

Algorithmic stability refers to the property of an algorithm being insensitive to perturbations of the input data, where the type of perturbation may vary depending on the setting. In this work, we develop a general framework to quantify the extent to which any ensembling strategy defined via averaging can yield stability guarantees for any type of data perturbation. Our main theoretical result is

This study develops a deep reinforcement learning framework for training Unmanned Aerial Vehicle (UAV) agents to navigate and monitor simulated wildfire environments. Results show that agents learn increasingly stable and effective behaviors over time, as demonstrated by converging loss trends, improved reward signals, and more consistent navigation patterns such as fire-boundary tracking. Overall

两名 AI 研究者因安全担忧离开 Anthropic 与谷歌：“房间里没有成年人” - NBC News

持续碾压对手的 AI 飓风模型 - Eye on the Tropics | Michael Lowry

Google DeepMind 发布 AlphaGenome 图谱，绘制 90 亿人类 DNA - Forbes

Anthropic 如何销毁图书以训练大语言模型：完整争议解读 - 36氪

Arlequin AI 融资 2800 万欧元，打造可大规模学习复杂关系的新型 AI 模型 - SiliconANGLE

ToolGrad：利用文本“梯度”高效生成工具使用数据集 - Google Research

基于大语言模型的模型无关个人身份信息（PII）检测 - Amazon Web Services (AWS)

手头有 5000 美元？现在就该买入的 3 只显而易见的人工智能（AI）股票 - Yahoo Finance

特朗普政府与 OpenAI 合作，为联邦雇员配备人工智能工具 - Fox Business

Amazon Quick 现已在桌面端正式上线 - Amazon Web Services (AWS)

中国与俄罗斯的不良行为者已在将 Anthropic 的 AI 武器化 - Politico

Anthropic 称已阻断可能利用 AI 研发生物武器的尝试 - CNN

Anthropic 称已阻断可能用于制造生物武器的尝试 - 《纽约时报》

病毒式传播的 AI 研究者警告“吓人至极”，克鲁兹称 - Politico