单词学习：![[57ddf6da-9581-47a2-acc6-e9e78ee78d36.mp3]]

新闻播报：![[b1c510ec-f1e2-44b8-9173-b1ab7f58b185.mp3]]

我国人工智能产业迈入万亿规模新阶段

简报：

• 根据工业和信息化部的数据，2023年我国人工智能产业的整体规模首次突破了万亿元人民币大关。

相关链接：

• 工信部：2023年人工智能产业规模首次突破万亿大关news/24063)

China's AI Industry Enters a New Trillion-Yuan Stage 🚀

Brief:

• According to data from the Ministry of Industry and Information Technology (MIIT), China's artificial intelligence industry's overall scale surpassed the one trillion RMB mark for the first time in 2023. 🥳

Related Links:

• MIIT: AI Industry Scale Exceeds One Trillion Yuan for the First Time in 2023

扎克伯格强推AI转型，Meta内部面临高压重组与裁员

简报：

• 为在AI领域追赶竞争对手，Meta CEO扎克伯格将公司带入“战时模式”，要求员工全力投入，并收缩元宇宙业务，将资源转向AI。
• 公司内部进行了激进的AI部门重组，成立超级智能实验室（MSL），但在数月内经历四次重组，引发了组织混乱和部分AI核心员工离职。
• Meta实施了更严格的绩效考核，要求管理者将15%至20%的员工评为“未达预期”，并裁掉了数千名被标记为低绩效的员工，加剧了内部紧张氛围。
• 前员工透露，公司内部存在组织僵化、愿景摇摆、新老团队摩擦等问题，高压环境导致员工普遍感到恐惧与不安全感。

相关链接：

• Meta 大逃杀！扎克伯格「地狱模式」曝光，不拼命搞 AI 就滚蛋
• 核心技术达国际先进水平，我国研发人工智能地震实时监测系统
• Meta 大逃杀！扎克伯格「地狱模式」曝光，不拼命搞 AI 就滚蛋
• 核心技术达国际先进水平，我国研发人工智能地震实时监测系统
• Meta 大逃杀！扎克伯格「地狱模式」曝光，不拼命搞 AI 就滚蛋

Zuckerberg Pushes Aggressive AI Transformation, Meta Faces High-Pressure Restructuring and Layoffs Internally

Briefing:

• To catch up with competitors in the AI sector, Meta CEO Mark Zuckerberg has plunged the company into 'wartime mode,' demanding full employee commitment and scaling back metaverse operations to redirect resources towards AI. 🤖
• The company underwent radical restructuring within its AI departments, establishing the Superintelligence Lab (MSL), but experienced four reorganizations within months, leading to organizational chaos and the departure of some core AI employees.
• Meta implemented stricter performance reviews, requiring managers to rate 15% to 20% of employees as 'falling short of expectations,' resulting in thousands of 'low-performing' employees being laid off and intensifying internal tension. 😓
• Former employees revealed issues such as organizational rigidity, wavering vision, and friction between new and old teams within the company. This high-pressure environment has led to widespread fear and insecurity among employees. 📉

Related Links:

• Meta 大逃杀！扎克伯格「地狱模式」曝光，不拼命搞 AI 就滚蛋
• 核心技术达国际先进水平，我国研发人工智能地震实时监测系统
• Meta 大逃杀！扎克伯格「地狱模式」曝光，不拼命搞 AI 就滚蛋
• 核心技术达国际先进水平，我国研发人工智能地震实时监测系统
• Meta 大逃杀！扎克伯格「地狱模式」曝光，不拼命搞 AI 就滚蛋

NVIDIA联手斯坦福推出通用游戏AI，并开源训练数据

简报：

• NVIDIA与斯坦福大学的研究人员合作，开源了一款能玩数千款2D视频游戏的通用AI智能体。
• 该项目同时公开了用于训练该AI的大规模数据集，包含超过4万小时的游戏视频。

相关链接：

• NVIDIA与斯坦福开源可“通玩”千款游戏的AI
• AI游戏智能体技术细节
• 公开的4万小时训练数据
• 通用AI在游戏领域的应用前景
• 项目开源公告

NVIDIA and Stanford Unveil General Gaming AI, Open-Sourcing Training Data 🎮

Briefing:

• NVIDIA has collaborated with researchers from Stanford University to open-source a general-purpose AI agent capable of playing thousands of 2D video games. 🧠
• The project also released a massive dataset used to train this AI, comprising over 40,000 hours of game video. 💾

Related Links:

• NVIDIA and Stanford Open-Source AI Capable of Playing Thousands of Games
• Technical Details of the AI Gaming Agent
• 40,000 Hours of Training Data Released
• Prospects of General AI Application in Gaming
• Project Open-Source Announcement

好的，这是根据您提供的资料总结的新闻简报：

Liquid AI发布2.6B参数模型，性能超越百亿级对手

简报：

• 边缘AI公司Liquid AI于圣诞节发布了仅有2.6B参数的实验性开源模型LFM2-2.6B-Exp。
• 该模型通过纯强化学习(RL)进行优化，在指令跟随能力基准测试中，表现超越了参数量为其263倍的DeepSeek R1-0528。
• 在数学推理（GSM8K）测试中，其得分超过82%，优于Llama3.2 3B和Gemma 3系列等模型。
• 该模型专为手机、笔记本等边缘设备设计，支持32K上下文长度，CPU推理速度是竞品的2倍。
• 模型权重已在Hugging Face平台完全开源，供开发者免费下载。

相关链接：

• LFM2-2.6B-Exp 模型地址

Liquid AI Unveils 2.6B Parameter Model, Outperforming Ten-Billion-Scale Rivals 🚀

Brief:

• Edge AI company Liquid AI released its experimental open-source model, LFM2-2.6B-Exp, featuring just 2.6 billion parameters, on Christmas Day.
• Optimized purely through Reinforcement Learning (RL), the model surpassed DeepSeek R1-0528, a model 263 times larger in parameter count, in instruction-following capability benchmarks.
• In mathematics reasoning (GSM8K) tests, it scored over 82%, outperforming models like Llama3.2 3B and the Gemma 3 series.
• Designed specifically for edge devices such as mobile phones and laptops 📱, the model supports a 32K context length and boasts CPU inference speeds twice as fast as competitors.
• The model weights are fully open-sourced on the Hugging Face platform ✨, available for free download by developers.

Related Link:

• LFM2-2.6B-Exp Model Link

证监会：将从三方面协同构建资本市场人工智能生态体系

简报： 中国证监会科技监管司副司长刘铁斌表示，面向“十五五”，将坚持促进创新与规范监管并重，从行业生态、机构生态和严守监管要求三方面协同发力，构建“合规有序、创新赋能、风险可控”的资本市场人工智能发展新格局。

相关链接：

• 证监会科技司刘铁斌：构建规范、健康的资本市场人工智能生态体系

CSRC: To Collaboratively Build Capital Market AI Ecosystem Across Three Fronts 📊

Brief: Liu Tiebin, Deputy Director of the Technology Supervision Department of the China Securities Regulatory Commission (CSRC), stated that, looking ahead to the "15th Five-Year Plan" period, the CSRC will adhere to promoting innovation while emphasizing standardized regulation. It will make concerted efforts from three aspects: industry ecology, institutional ecology, and strict adherence to regulatory requirements, to build a new landscape for capital market artificial intelligence development that is "compliant and orderly, innovation-enabled, and risk-controllable." 🚀✨

Related Link:

• CSRC Technology Department's Liu Tiebin: Building a Standardized and Healthy Capital Market AI Ecosystem

前主管投靠对手后，台积电全面更换先进工艺代号防泄密

简报：

• 为防止前员工罗唯仁投靠英特尔后窥探公司2奈米以下制程发展，台积电已全面升级防谍措施。
• 具体措施包括将2奈米至1奈米的先进工艺专案代号，以及重要客户的专案代号全部进行更换。
• 同时，公司也强化了供应链的保密机制，并采取专厂生产等方式，以确保核心技术不外流，巩固其行业龙头地位。

相关链接：

• 台積電防技術外洩再出招！　換專案代號、專廠生產

TSMC Completely Changes Advanced Process Codes to Prevent Leaks After Former Executive Defects to Rival 🛡️

Briefing:

• To prevent former employee Lo Wei-jen from prying into the company's sub-2nm process development after joining Intel, TSMC has fully upgraded its anti-espionage measures.
• Specific measures include changing all project codes for advanced processes from 2nm down to 1nm, as well as project codes for important clients.
• Concurrently, the company has strengthened its supply chain's confidentiality mechanisms and adopted dedicated factory production, among other methods, to ensure core technologies do not leak out and to solidify its industry-leading position. 🔒

Related Link:

• 台積電防技術外洩再出招！　換專案代號、專廠生產

联想发布AI超级智能体，整合PC与手机体验

简报：

• 联想正式公布其战略级AI——“超级智能体”（AI Agent），作为其AI PC的核心体验。
• 该AI的核心目标是打通个人电脑（PC）与手机生态系统，实现跨设备的无缝协同。
• “超级智能体”将作为用户的个人AI助理，能够基于在本地运行的个人大模型和个人知识库，理解用户意图并主动完成任务。
• 此举是联想在其“AI for All”战略下，针对AI PC时代的重要产品布局，旨在重塑人机交互体验。

相关链接：

• 联想集团执行副总裁兼中国区总裁刘军：联想“超级智能体”是AI PC的核心体验
• 联想发布首款AI PC，售价8999元起
• 联想曝光战略级AI“超级智能体” 将打通PC与手机生态

Lenovo Unveils AI Super Agent, Integrating PC and Mobile Experiences

Brief:

• Lenovo has officially announced its strategic AI – the "AI Super Agent" – as the core experience for its AI PCs. 🧠
• The AI's core objective is to bridge the gap between PC and mobile ecosystems, enabling seamless cross-device collaboration. 📱💻
• The "AI Super Agent" will function as a personal AI assistant for users, capable of understanding user intent and proactively completing tasks based on locally run personal large models and knowledge bases.
• This move is a significant product strategy under Lenovo's "AI for All" initiative for the AI PC era, aiming to redefine human-computer interaction.

Related Links:

• Liu Jun, Executive Vice President and President of Lenovo China: Lenovo's 'AI Super Agent' is the Core Experience of AI PCs
• Lenovo Releases Its First AI PC, Starting from 8999 RMB
• Lenovo Unveils Strategic AI 'Super Agent' to Integrate PC and Mobile Ecosystems