单词学习：![[2a88f84a-b8bd-4370-96a1-3e24c18daf76.mp3]]

新闻播报：![[c2ce178f-5aa9-46a1-9881-18e75b8fb010.mp3]]

摩尔线程发布新GPU架构“花港”，并推出“庐山”芯片及AI笔记本

简报：

• 摩尔线程于12月20日发布新一代GPU架构“花港”，其算力密度提升50%，能效提升10倍，并支持10万卡以上规模的智算集群方案。
• 基于“花港”架构，摩尔线程推出了“庐山”高性能GPU芯片，其AI性能可提升64倍，光线追踪性能提升50倍。
• 大会上还发布了AI算力笔记本MTT AIBOOK，搭载自主研发的“长江”SoC芯片，异构AI算力达50TOPS，预售价为9999元。

相关链接：

• 摩尔线程发布新一代GPU架构“花港”
• 摩尔线程发布“庐山”GPU芯片，AI性能提升64倍
• 摩尔线程发布AI算力笔记本MTT AIBOOK

Moore Threads Unveils New GPU Architecture "Huguang", Introduces "Lushan" Chip and AI Laptop 🚀

Briefing:

• Moore Threads announced its new-generation GPU architecture "Huguang" on December 20th, boasting a 50% increase in compute density, a 10x improvement in energy efficiency, and support for intelligent computing clusters with over 100,000 cards.
• Based on the "Huguang" architecture, Moore Threads launched the "Lushan" high-performance GPU chip, which offers a 64x boost in AI performance and a 50x increase in ray tracing performance. 🎮
• The conference also saw the release of the MTT AIBOOK AI computing laptop, powered by the self-developed "Changjiang" SoC chip, delivering heterogeneous AI computing power of 50 TOPS, with a pre-sale price of 9,999 RMB. 💻

Related Links:

• Moore Threads Releases New Generation GPU Architecture "Huguang"
• Moore Threads Launches "Lushan" GPU Chip, AI Performance Boosted by 64x
• Moore Threads Releases AI Computing Laptop MTT AIBOOK

主打百倍能效，模拟AI芯片商Mythic获1.25亿美元融资

简报：

• AI芯片公司Mythic完成新一轮1.25亿美元融资，由深科技风险投资公司DCVC领投，资金将用于产品开发、软件完善及商业扩张。
• 公司计划开发能效比现有GPU高出100倍的模拟处理单元（APU），主要应用于数据中心、汽车系统、机器人及国防领域。
• Mythic的芯片采用模拟内存计算技术，宣称其架构目前每瓦可提供120万亿次操作，并能高效运行万亿参数的大型语言模型。
• 公司CEO强调，其APU产品并非旨在取代GPU，而是作为GPU旁的加速器与其形成互补关系。

相关链接：

• Mythic 融资 1.25 亿美元，力图以超高能效 AI 芯片挑战 NVIDIA

Analog AI Chipmaker Mythic Secures dJyH6kDkfC5RlM16.answerText25 Million Funding, Targeting 100x Energy Efficiency

Brief:

• AI chip company Mythic has closed a new dJyH6kDkfC5RlM16.answerText25 million funding round, led by deep-tech venture capital firm DCVC. The capital will be allocated towards product development, software enhancement, and business expansion. 🚀
• The company plans to develop Analog Processing Units (APUs) that boast 100 times the energy efficiency of existing GPUs, primarily targeting data centers, automotive systems, robotics, and defense sectors.
• Mythic's chips leverage analog in-memory computing technology, with the company claiming its architecture currently delivers 120 trillion operations per watt and can efficiently run large language models with trillions of parameters. 💡
• The company's CEO emphasized that their APU products are not designed to replace GPUs but rather to complement them as accelerators.

Related Links:

• Mythic Raises dJyH6kDkfC5RlM16.answerText25 Million, Aiming to Challenge NVIDIA with Ultra-High Efficiency AI Chips 🔗

ChatGPT迎来个性化更新，聊天风格可自定义

简报：

• OpenAI为ChatGPT网页端及App推出更新，允许用户自行调节AI的亲和力与暖心程度，选项覆盖“温和体贴”与“热情洋溢”两个维度。
• 用户可在自定义面板中为GPT选择多种性格，包括“专业可靠”、“亲和友善”、“直言不讳”、“天马行空”、“吐槽达人”等。
• 更新还支持调整GPT使用表情符号、标题和列表的频率。
• 另一项新功能允许用户直接在聊天界面中对ChatGPT生成的电子邮件文本进行修改、排版或对选中部分进行特定修改。

相关链接：

• ChatGPT 新增暖心程度调节功能，用户可自定义 AI 聊天风格
• macOS 版 ChatGPT 应用官宣 2026 年 1 月停用 AI 语音模式

ChatGPT Unleashes Personalization Updates, Customize Your AI's Chat Style! 🤖

Briefing:

• OpenAI has rolled out updates for both the ChatGPT web client and app, empowering users to fine-tune the AI's friendliness and warmth, with options ranging from "Gentle & Considerate" to "Enthusiastic & Vibrant." 💖
• Users can now select from a variety of personalities for GPT within the customization panel, including "Professional & Reliable," "Friendly & Approachable," "Frank & Direct," "Imaginative & Unconventional," and "Sarcastic Critic."
• The update also introduces controls for adjusting GPT's frequency of using emojis, headings, and lists.
• A new feature allows users to directly modify, format, or make specific adjustments to selected portions of ChatGPT-generated email text right within the chat interface. ✍️

Related Links:

• ChatGPT Adds Warmth Adjustment Feature, Users Can Customize AI Chat Style
• macOS ChatGPT App Officially Announces Discontinuation of AI Voice Mode in January 2026

AI巨头加强未成年人保护，将上线年龄预测功能

简报：

• OpenAI与Anthropic宣布将通过AI模型预测用户年龄，以加强对未成年用户的识别与保护。
• OpenAI已更新其“模型规范”，在与13至17岁青少年互动时将优先考虑安全，并正在开发一个年龄预测模型，在检测到用户可能未满18岁时自动启用保护机制。
• Anthropic的政策禁止18岁以下用户使用其产品，目前正开发一套检测系统，旨在通过对话中的“微妙迹象”识别并自动关停未成年人账号。
• 此举是在立法者就AI对心理健康的影响向相关公司施压的背景下推出的。

相关链接：

• 防范未成年人风险
  与Anthropic将推出AI年龄预测功能
• 为防范未成年人风险，OpenAI、Anthropic将推出AI年龄预测功能

AI Giants Boost Minor Protection, Set to Launch Age Prediction Features 🛡️🧒

Brief:

• OpenAI and Anthropic have announced they will use AI models to predict user age, enhancing identification and protection for underage users.
• OpenAI has updated its "Model Specifications," prioritizing safety when interacting with teenagers aged 13 to 17. It is also developing an age prediction model that will automatically activate protective mechanisms if a user is detected as potentially under 18.
• Anthropic's policy prohibits users under 18 from using its products. The company is currently developing a detection system designed to identify and automatically shut down underage accounts by recognizing "subtle cues" in conversations.
• This move comes amid pressure from lawmakers on relevant companies regarding AI's impact on mental health. ⚠️

Related Links:

• 防范未成年人风险
  与Anthropic将推出AI年龄预测功能
• 为防范未成年人风险，OpenAI、Anthropic将推出AI年龄预测功能

Kimi模型将集成至微软Office，实现产品自动化

简报：

• 继微软Azure月初宣布接入月之暗面（Moonshot AI）的Kimi k2模型后，双方合作正向应用层拓展。
• 据披露，微软计划于2025年12月上线全新的Agent功能，该功能将深度集成Kimi模型。
• 此举旨在利用Kimi的长程推理和工具调用能力，实现Office产品的全面自动化。

相关链接：

• 微软联手 Kimi 打造 Office 自动化 Agent，AI 云巨头开启本土化“自我修正”
• 微软 Win11 版 Copilot 应用上线实时文本编辑，用户动动嘴就能改稿

Kimi Model to be Integrated into Microsoft Office, Enabling Product Automation

Briefing:

• Following Microsoft Azure's announcement earlier this month to integrate Moonshot AI's Kimi k2 model, the collaboration between the two parties is now expanding to the application layer. ✨
• It has been revealed that Microsoft plans to launch a new Agent feature in December 2025, which will deeply integrate the Kimi model.
• This initiative aims to leverage Kimi's long-context reasoning and tool-calling capabilities to achieve comprehensive automation across Office products. 🚀

Related Links:

• 微软联手 Kimi 打造 Office 自动化 Agent，AI 云巨头开启本土化“自我修正”
• 微软 Win11 版 Copilot 应用上线实时文本编辑，用户动动嘴就能改稿

谷歌Gemini应用上线新功能，可验证视频是否为AI生成

简报：

• 谷歌在其Gemini应用中推出了一项全新的验证功能，旨在帮助用户检查视频内容是否由谷歌自家的人工智能技术生成。
• 该功能旨在提升平台内容的透明度，让用户能更清楚地了解所接触内容的来源和真实性，从而做出更明智的决策。
• 除了验证功能，Gemini应用还提供了多种AI工具，以增强用户在视频创作与编辑方面的体验。
• 目前，新发布的Gemini 3 Flash模型已成为Gemini应用中的默认模型，为包括新功能在内的日常任务提供支持。

相关链接：

• 谷歌 Gemini 应用推出视频生成 AI 检测新功能
• 持續進化！Google推Gemini 3 Flash　強大功能一次看

Google Gemini App Launches New Feature to Verify If Videos Are AI-Generated

Briefing:

• Google has rolled out a brand-new verification feature within its Gemini app, designed to help users check whether video content was generated by Google's own artificial intelligence technology. 🤖
• This feature aims to enhance content transparency on the platform, empowering users to better understand the origin and authenticity of the content they encounter, leading to more informed decisions. ✨
• Beyond the verification tool, the Gemini app also offers various AI tools to enrich the user experience in video creation and editing. 🎥
• Currently, the newly released Gemini 3 Flash model has become the default model in the Gemini app, supporting daily tasks, including the new features.

Related Links:

• Google Gemini App Launches New AI Detection Feature for Video Generation
• Continually Evolving! Google Rolls Out Gemini 3 Flash, Powerful Features at a Glance