单词学习：![[bfa74acc-7e61-4e41-a40a-e0ef96acd4b2.mp3]]

新闻播报：![[5e4b9999-c283-47fe-af86-a6f2e071012d.mp3]]

联合国设立两大新机制，加强全球人工智能治理

简报：

• 联合国大会于8月26日通过决议，决定设立新的全球人工智能治理机制，以促进可持续发展和弥合数字鸿沟。
• 将设立一个由40名专家组成的“人工智能独立国际科学小组”，负责发布基于证据的科学评估，分析人工智能的机遇、风险和影响。
• 同时将启动“人工智能治理全球对话”机制，供各国政府及其他利益相关方就国际合作、分享最佳实践等内容展开年度对话。
• 此举旨在应对成员国对人工智能飞速发展可能给民主及人权带来威胁的担忧。

相关链接：

• 新华社报道
• 36氪快讯
• AI基地报道

UN Establishes Two New Mechanisms to Strengthen Global AI Governance

Brief:

• The United Nations General Assembly adopted a resolution on August 26, deciding to establish new global artificial intelligence governance mechanisms to promote sustainable development and bridge the digital divide. 🌍
• An 'Independent International Scientific Panel on Artificial Intelligence' comprising 40 experts will be established. Its mandate will be to publish evidence-based scientific assessments analyzing the opportunities, risks, and impacts of AI.
• Concurrently, a 'Global Dialogue on AI Governance' mechanism will be launched, allowing governments and other stakeholders to engage in annual discussions on international cooperation, sharing best practices, and more. 🗣️
• This initiative aims to address member states' concerns regarding the potential threats that the rapid development of artificial intelligence could pose to democracy and human rights.

Related Links:

• Xinhua News Agency Report
• 36Kr Newsflash
• AI Base Report 💡

资本热捧AI医疗沟通，初创公司四个月估值达75亿

简报：

• 专注于患者沟通自动化的AI医疗初创公司Assort Health，在完成A轮融资仅4个月后，估值从不到25亿人民币飙升至75亿人民币，涨幅超200%。
• 该公司在由光速创投领投的B轮融资中筹集了约5000万美元，其核心产品是AI语音代理系统，用于处理医疗机构的预约安排、取消等前台重复性工作。
• Assort Health主要服务于中小型专科医疗机构，目前年度经常性收入略高于300万美元，并已将业务从骨科扩展至妇产科、皮肤科和牙科等领域。
• 此轮融资反映了AI医疗助手赛道的投资热潮，近期同赛道的EliseAI和Hello Patient也分别完成了大额融资。

相关链接：

• AI医疗助手赛道彻底爆发！这家初创公司4个月估值狂飙3倍达75亿

Capital Rushes into AI Healthcare Communication, Startup Valued at 7.5 Billion RMB ($1 Billion USD) in Four Months

Brief:

• Assort Health, an AI healthcare startup specializing in patient communication automation, saw its valuation skyrocket by over 200% from less than 2.5 billion RMB to 7.5 billion RMB ($1 billion USD) in just four months after completing its Series A funding round. 🚀
• The company raised approximately $50 million in its Series B funding round led by Lightspeed Venture Partners. Its core product is an AI voice agent system designed to handle repetitive front-office tasks for medical institutions, such as appointment scheduling and cancellations. 🗣️
• Assort Health primarily serves small and medium-sized specialty medical institutions, currently generating just over $3 million in annual recurring revenue (ARR). It has also expanded its services from orthopedics to fields such as obstetrics and gynecology, dermatology, and dentistry. 🏥
• This funding round reflects the intense investment frenzy in the AI medical assistant sector, with peers EliseAI and Hello Patient also recently completing significant funding rounds.

Related Link:

• AI Medical Assistant Sector Explodes! This Startup's Valuation Skyrockets 3x to 7.5 Billion in 4 Months

字节发布AI视频模型Waver 1.0，支持多风格与连贯叙事

简报：

• 字节跳动发布了一款名为Waver 1.0的AI视频生成模型，支持文本到视频和图像到视频的转换。
• 该模型能够生成长达10秒的视频，并支持包括极致现实、动画、粘土、毛绒等多种艺术风格。
• Waver 1.0具备多镜头叙事能力，能在镜头切换时保持主题、风格和氛围的高度一致性。
• 根据人工评估，该模型在运动质量、视觉质量和提示跟随性等方面的表现优于现有的开源和闭源模型。

相关链接：

• 字节的新AI视频模型 Waver 1.0 发布，支持多种艺术风格的视频生成
• Waver 1.0 项目官网

ByteDance Unveils AI Video Model Waver 1.0, Supporting Multiple Styles and Coherent Storytelling

Summary:

• ByteDance has released an AI video generation model named Waver 1.0, which supports both text-to-video and image-to-video conversion. 🚀
• The model can generate videos up to 10 seconds long and supports a variety of artistic styles, including hyperrealism, animation, claymation, and plush. ✨
• Waver 1.0 features multi-shot storytelling capabilities, maintaining high consistency in theme, style, and atmosphere during scene transitions.
• According to human evaluations, the model outperforms existing open-source and closed-source models in terms of motion quality, visual quality, and prompt adherence. 🌟

Related Links:

• ByteDance's New AI Video Model Waver 1.0 Released, Supporting Multi-Style Video Generation
• Waver 1.0 Project Official Website

谷歌翻译融入Gemini AI，推出实时同传与语言陪练新功能

简报：

• 谷歌翻译基于其Gemini AI模型进行了重大升级，推出了实时同传（Live Translate）和智能语言陪练（Practice）两项新功能。
• 实时同传功能可为对话提供无缝的实时语音翻译和屏幕字幕，能自动识别语音停顿和语调，已于2025年8月26日起在美国、印度和墨西哥上线，支持超过70种语言。
• 智能语言陪练功能可以根据用户的学习目标生成个性化的互动场景（如酒店入住、餐厅点餐），并提供实时反馈以提升口语能力。
• 语言陪练功能目前处于测试阶段，初期支持英语用户练习西班牙语和法语，以及西班牙语、法语和葡萄牙语用户练习英语。

相关链接：

• 谷歌翻译大升级，实时同传+AI陪练，学外语像聊天一样简单

Google Translate Integrates Gemini AI, Unveils Real-time Live Translate and Language Practice Features

Briefing:

• Google Translate has received a significant upgrade leveraging its Gemini AI model, introducing two powerful new features: Live Translate and intelligent Language Practice. 🚀
• The Live Translate feature provides seamless real-time voice translation and on-screen captions for conversations. It automatically recognizes speech pauses and intonation. This feature launched on August 26, 2025, in the US, India, and Mexico, supporting over 70 languages.
• The intelligent Language Practice feature can generate personalized interactive scenarios (e.g., hotel check-in, ordering food at a restaurant) based on users' learning goals, and provides real-time feedback to improve speaking proficiency. 🗣️
• The Language Practice feature is currently in its beta phase. Initially, it supports English speakers practicing Spanish and French, as well as Spanish, French, and Portuguese speakers practicing English. 🌍

Related Links:

• Google Translate's Big Upgrade: Real-time Live Translate + AI Practice Makes Language Learning a Breeze

谷歌Gemini 2.5 Flash发布，支持文字指令精准修图

简报：

• 谷歌DeepMind推出了Gemini 2.5 Flash图像编辑模型，用户可通过文字指令对图片进行高精度修改，且在多项任务中表现优于GPT-4o。
• 新模型支持“角色一致性”功能，即使在姿势、背景或光线变化时，也能在多张图像中保持人物或物体的外观一致。
• 用户无需手动圈选即可进行局部编辑，如背景虚化、移除物体等，并支持最多三张图像的融合以及“风格迁移”。
• 该模型具备“现实推理”功能，可模拟简单的因果关系，如气球飞向仙人掌及之后的结果。
• 用户可在Gemini应用中将模型切换至“Flash”进行体验，生成的图像会附带可见及不可见的SynthID数字水印。

相关链接：

• 谷歌推出全新的Gemini 2.5 Flash图像编辑模型 精准编辑轻松搞定修图
• 谷歌 Gemini 2.5 Flash 升级 AI 修图功能，多项表现优于 GPT-4o

Google Gemini 2.5 Flash Released, Enables Precise Image Editing via Text Commands 📸

Brief:

• Google DeepMind has launched the Gemini 2.5 Flash image editing model, allowing users to make high-precision modifications to images via text commands. It also outperforms GPT-4o in multiple tasks.
• The new model supports a "character consistency" feature, maintaining the consistent appearance of people or objects across multiple images, even with changes in pose, background, or lighting. ✨
• Users can perform local edits without manual selection, such as background blurring or object removal. It also supports merging up to three images and "style transfer."
• The model possesses "realistic reasoning" capabilities, simulating simple cause-and-effect relationships, such as a balloon flying towards a cactus and its subsequent outcome.
• Users can experience the model by switching to "Flash" in the Gemini app. Generated images will come with both visible and invisible SynthID digital watermarks. 💧

Related Links:

• Google Launches New Gemini 2.5 Flash Image Editing Model, Making Precise Edits Easy
• Google Gemini 2.5 Flash Upgrades AI Photo Editing, Outperforming GPT-4o in Multiple Aspects

首款AI勒索软件PromptLock曝光，利用OpenAI模型跨平台攻击

简报：

• 网络安全公司ESET报告发现全球首个AI勒索软件，并将其命名为PromptLock。
• 该软件滥用OpenAI开源的gpt-oss:20b语言模型，在受感染设备上本地生成恶意Lua代码。
• PromptLock具备跨平台能力，可在Windows、Linux和macOS系统上执行搜索、窃取和加密文件的操作。
• 安全专家警告称这是恶意利用本地AI的早期信号，OpenAI已回应称正采取措施降低模型被利用的风险。

相关链接：

• 全球首现AI勒索软件PromptLock：跨平台黑客新武器来袭
• 全球首个 AI 勒索软件现身：滥用 OpenAI 开源模型，跨平台搜索、窃取、加密文件

PromptLock: The First AI Ransomware Exposed, Exploiting OpenAI Models for Cross-Platform Attacks

Briefing:

• Cybersecurity firm ESET reports the discovery of the world's first AI ransomware, naming it PromptLock. 🤖
• The software abuses OpenAI's open-source gpt-oss:20b language model to locally generate malicious Lua code on infected devices.
• PromptLock boasts cross-platform capabilities, able to search, steal, and encrypt files on Windows, Linux, and macOS systems.
• Security experts warn this is an early signal of malicious local AI exploitation. OpenAI has responded, stating it is taking measures to mitigate the risk of its models being misused. 🚨

Related Links:

• PromptLock: The World's First AI Ransomware, a New Cross-Platform Hacking Weapon Appears
• World's First AI Ransomware Emerges: Abuses OpenAI Open-Source Model to Search, Steal, and Encrypt Files Cross-Platform 💻

马斯克起诉苹果与OpenAI，指控其AI合作构成垄断

简报：

• 特斯拉与xAI创始人马斯克已在美国德州沃斯堡联邦法院正式对苹果公司与OpenAI提起诉讼。
• 诉讼指控苹果在其“Apple Intelligence”中深度集成ChatGPT，构成了对AI入口的“垄断”，此举边缘化了其他AI产品，并直接损害了马斯克旗下的Grok与X应用。
• 马斯克在诉讼中要求数额达数十亿美元的赔偿，并请求法院下令禁止苹果与OpenAI继续执行该“非法协议”。
• OpenAI回应称马斯克的指控是“持续骚扰的一部分”，而苹果公司暂未公开回应。

相关链接：

• 首富的攪屎棍！蘋果綁ChatGPT惹怒馬斯克　一次性怒告2巨頭蘋果、OpenAI

Musk Sues Apple and OpenAI, Alleging AI Partnership Constitutes Monopoly

Brief:

• Elon Musk, founder of Tesla and xAI, has officially filed a lawsuit against Apple Inc. and OpenAI in the U.S. Federal Court in Fort Worth, Texas. ⚖️
• The lawsuit alleges that Apple's deep integration of ChatGPT into its 'Apple Intelligence' constitutes a 'monopoly' over the AI entry point, marginalizing other AI products and directly harming Musk's Grok and X applications. 😠
• Musk is seeking billions of dollars in damages in the lawsuit and has requested the court to issue an order prohibiting Apple and OpenAI from continuing with this 'illegal agreement'.
• OpenAI has responded, stating that Musk's accusations are 'part of an ongoing harassment campaign,' while Apple Inc. has not yet publicly commented. 📉

Related Link:

• Billionaire's Troublemaker! Apple's ChatGPT Tie-up Enrages Musk, Sues Two Giants Apple and OpenAI at Once

輝達發布機器人超級大腦Jetson Thor，引爆概念股漲停潮

簡報：

• 輝達(NVIDIA)於8月25日正式發布被稱為「機器人新大腦」的Jetson AGX Thor開發者套件與生產模組，並已全面上市。
• Jetson Thor採用NVIDIA Blackwell GPU架構，AI運算能力是前代Jetson Orin的7.5倍，開發者套件定價3,499美元起，鎖定人形機器人、製造、物流等高階應用。
• 此消息直接帶動台灣股市機器人概念股於8月26日集體上漲，其中所羅門（2359）、昆盈（2365）等多檔股票攻上涨停。
• 輝達同時也推出用於自動駕駛的DRIVE AGX Thor平台，已獲比亞迪、小米、Volvo等11家全球車廠與自駕貨運公司採用。
• Amazon Robotics、Boston Dynamics、Figure等國際企業已率先採用Jetson Thor平台；市場研究機構TrendForce預估，至2028年相關晶片市場規模可突破4800萬美元。

相關連結：

• 盤中／黃仁勳親揭機器人新大腦Jetson Thor！概念股所羅門又漲到鎖門了
• 價格好甜！輝達Jetson Thor開發者套件上市　Amazon來點讚
• 新聞幕後／黃仁勳送機器人「新大腦」真值錢！劍指4800萬美元市場
• 輝達推自駕大腦DRIVE Thor套件　11家全球車廠全跟進！

NVIDIA Unleashes Robot Superbrain Jetson Thor, Igniting a Surge in Concept Stocks 🚀

Briefing:

• NVIDIA officially launched its Jetson AGX Thor developer kit and production module, dubbed the "new brain for robots," on August 25th, and it is now fully available.
• Jetson Thor utilizes NVIDIA's Blackwell GPU architecture, boasting an AI computing capability 7.5 times that of its predecessor, Jetson Orin. The developer kit starts at $3,499, targeting high-end applications such as humanoid robots, manufacturing, and logistics.
• This announcement directly triggered a collective rise in Taiwan's robotics concept stocks on August 26th, with several shares, including Solomon (2359) and KYE Systems (2365), hitting their daily trading limits. 📈
• NVIDIA also introduced the DRIVE AGX Thor platform for autonomous driving, which has already been adopted by 11 global automakers and autonomous freight companies, including BYD, Xiaomi, and Volvo.
• International companies such as Amazon Robotics, Boston Dynamics, and Figure have already begun adopting the Jetson Thor platform. Market research firm TrendForce estimates that the related chip market size could exceed $48 million by 2028. 💰

Related Links:

• 盤中／黃仁勳親揭機器人新大腦Jetson Thor！概念股所羅門又漲到鎖門了
• 價格好甜！輝達Jetson Thor開發者套件上市　Amazon來點讚
• 新聞幕後／黃仁勳送機器人「新大腦」真值錢！劍指4800萬美元市場
• 輝達推自駕大腦DRIVE Thor套件　11家全球車廠全跟進！