单词学习：![[1700081e-0c9f-498f-9809-f80759933dc5.mp3]]

ChatGPT在韩月活用户破2000万，占智能手机用户四成

简报：

• 数据显示，2025年8月，韩国ChatGPT月活跃用户数首次突破2000万，达到创纪录的2031万，远高于一年前的407万。
• 该用户数约占韩国5012万智能手机用户的40%。
• 从用户构成来看，男性占50.1%，女性占49.9%，比例基本持平；年龄在20至40岁之间的用户占据了近70%。

ChatGPT's MAU in Korea Exceeds 20 Million, Accounts for 40% of Smartphone Users

Briefing:

• Data shows that in August 2025, ChatGPT's Monthly Active Users (MAU) in South Korea surpassed 20 million for the first time, reaching a record high of 20.31 million, significantly up from 4.07 million a year ago. 🚀
• This user base accounts for approximately 40% of South Korea's 50.12 million smartphone users. 📱
• In terms of user demographics, males constitute 50.1% and females 49.9%, indicating a roughly even split; users aged between 20 and 40 years old make up nearly 70%. 📊

腾讯发布开源框架Youtu-Agent，两步搭建专属智能体

简报：

• 腾讯优图实验室正式开源智能体框架Youtu-Agent，旨在为开发者提供高效、易用的智能体开发工具。
• 该框架以“开箱即用”为理念，用户通过拉取代码和编写YAML配置文件两步，即可快速搭建和运行智能体。
• Youtu-Agent在WebWalkerQA和GAIA等权威测试中表现优异，不依赖昂贵的闭源模型，支持二次开发和定制化。
• 框架面向文件管理、数据分析、学术研究、信息综述等实际应用场景，并已为腾讯云多个产品业务提供支持。

相关链接：

• Youtu-Agent Github项目地址

Tencent Launches Open-Source Framework Youtu-Agent: Build Your Own AI Agent in Two Steps ✨

Briefing:

• Tencent Youtu Lab has officially open-sourced its AI agent framework, Youtu-Agent, aiming to provide developers with efficient and easy-to-use agent development tools.
• Based on an "out-of-the-box" philosophy, users can quickly set up and run agents in just two steps: pulling the code and writing a YAML configuration file. 🚀
• Youtu-Agent has demonstrated excellent performance in authoritative tests like WebWalkerQA and GAIA. It does not rely on expensive closed-source models and supports secondary development and customization. 💡
• The framework targets practical application scenarios such as file management, data analysis, academic research, and information summarization, and has already provided support for multiple Tencent Cloud product businesses.

Related Link:

• Youtu-Agent Github Project Link

苏宁携手寒武纪，灵思大模型实现效能与成本双突破

简报：

• 苏宁易购发布其灵思大模型，该模型通过与国产寒武纪算力平台深度协作，实现了推理效率提升30%和训练成本降低70%的双重优化。
• 灵思大模型参数规模覆盖3B至700B，具备多模态能力，旨在推动算力自主可控。
• 目前，该模型已在苏宁易购的物流供应链、零售营销等12个核心场景中落地应用。

相关链接：

• 苏宁打响国产AI反击战：灵思大模型携手寒武纪算力平台，推理效率暴涨30%训练成本狂降70%

Suning Partners with Cambricon: Lingsi Large Model Achieves Dual Breakthroughs in Efficiency and Cost 🚀💡

Brief:

• Suning.com has launched its Lingsi Large Model, which, through deep collaboration with the domestic Cambricon computing platform, has achieved a dual optimization: a 30% increase in inference efficiency and a 70% reduction in training costs. 💰
• The Lingsi Large Model boasts parameter sizes ranging from 3B to 700B and possesses multimodal capabilities, aiming to promote independent and controllable computing power.
• Currently, the model has been implemented and applied across 12 core scenarios at Suning.com, including logistics supply chain and retail marketing.

Related Link:

• Suning's Domestic AI Counterattack: Lingsi Large Model Teams Up with Cambricon Computing Platform, Inference Efficiency Soars by 30%, Training Costs Plummet by 70%

微信响应新规，将为AI生成内容添加标识

简报：

• 微信公众平台于2025年9月1日发布公告，为响应《人工智能生成合成内容标识办法》要求，将对AI生成合成的内容添加显式和隐式标识。
• 平台将通过提供官方标识提示和要求用户主动声明两种形式，对AI生成内容进行标注，以保障用户获取信息的透明度与可信度。
• 标识分为显式（如文字、图形）和隐式（通过技术措施嵌入文件数据）两种，平台也会对可能是AI生成合成的内容进行相应提示。
• 《人工智能生成合成内容标识办法》由国家网信办等多部门联合制定，已于2025年9月1日正式施行。

相关链接：

• 微信：平台将对 AI 生成合成内容添加显式和隐式标识
• DeepSeek：已在平台内对 AI 生成合成内容添加标识，用户不得恶意删除

WeChat to Label AI-Generated Content Following New Regulations

Brief:

• On September 1, 2025, WeChat Official Accounts Platform announced that, in response to the requirements of the "Measures for the Labeling of AI-Generated Synthetic Content," it will add both explicit and implicit labels to AI-generated synthetic content. 🤖
• The platform will label AI-generated content through two methods: providing official labeling prompts and requiring users to actively declare, thereby ensuring the transparency and credibility of information for users. ✨
• Labels will be categorized as explicit (e.g., text, graphics) and implicit (embedded in file data via technical measures). The platform will also provide corresponding prompts for content that may potentially be AI-generated synthetic. 🏷️
• The "Measures for the Labeling of AI-Generated Synthetic Content," jointly formulated by the Cyberspace Administration of China (CAC) and multiple other departments, officially came into effect on September 1, 2025.

Related Links:

• WeChat: Platform to Add Explicit and Implicit Labels to AI-Generated Synthetic Content
• DeepSeek: Labels Added to AI-Generated Synthetic Content on Platform, Users Prohibited from Maliciously Deleting

苹果AI新动向：内部聊天机器人“Asa”曝光，用于培训零售员工

简报：

• 苹果正在内部测试一款名为“Asa”的AI聊天机器人，旨在帮助Apple Store零售员工快速掌握产品知识，从而提升与顾客的互动能力。
• 该工具已被整合进苹果内部专为员工教育培训和销售辅助设计的“SEED”应用服务中。
• 与竞争对手直接推出消费者AI服务不同，苹果选择先从内部零售场景引入AI，尚未发布面向消费者的通用型聊天机器人。
• 此举被视为苹果谨慎AI战略的一部分，即先在内部验证AI的价值和可靠性，再逐步将成熟技术推向市场。

相关链接：

• 苹果内部AI聊天机器人“Asa”曝光，为零售员工打造专属AI
• 苹果 FastVLM 模型开放试用：Mac 用户秒享“闪电级”视频字幕

Apple's New AI Move: Internal Chatbot "Asa" Revealed for Retail Staff Training 🤖

Brief:

• Apple is internally testing an AI chatbot named "Asa," designed to help Apple Store retail employees quickly master product knowledge and enhance their interaction capabilities with customers. 🛍️
• This tool has been integrated into Apple's internal "SEED" application service, specifically designed for employee education, training, and sales assistance.
• Unlike competitors who directly launch consumer-facing AI services, Apple has opted to introduce AI starting with internal retail scenarios and has not yet released a general-purpose chatbot for consumers.
• This move is seen as part of Apple's cautious AI strategy: first validating the value and reliability of AI internally, then gradually bringing mature technology to the market. ✨

Related Links:

• Apple's Internal AI Chatbot "Asa" Revealed, Custom-Built for Retail Employees
• Apple FastVLM Model Open for Trial: Mac Users Instantly Enjoy "Lightning-Fast" Video Subtitles