单词学习：![[5b0fc808-26cc-484a-99bf-fb98700893dc.mp3]]

新闻播报：![[27d19591-8c8e-44d6-809a-7cecba1b3b56.mp3]]

信通院报告：2024年我国AI产业规模突破9000亿元

简报：

• 中国信息通信研究院测算，2024年我国人工智能产业规模已超9000亿元，同比增长24%。
• 截至2025年9月，我国人工智能企业数量超过5300家，占全球总数的15%，已形成完整的产业体系。
• 该测算成果于9月23日在2025人工智能产业及赋能新型工业化大会上发布。

相关链接：

• 中国信通院：2024 年我国人工智能产业规模超 9000 亿元，同比增长 24%
• 阿里巴巴宣布与英伟达开展 Physical AI 合作，推动具身智能应用落地
• 阿里巴巴 CEO 吴泳铭：正在推进三年 3800 亿的 AI 基础设施建设计划
• 阿里云栖大会一口气发布千问 3-VL、万相 2.5 等六大模型 + 通义百聆新品牌，覆盖文本、视觉、语音、视频、代码、图像全场景
• 谷歌最强生图 AI 模型加持：Mixboard 可视化你的创意，开放画布 + 自然语言编辑

CAICT Report: China's AI Industry Scale Exceeds 900 Billion Yuan in 2024

Brief:

• According to calculations by the China Academy of Information and Communications Technology (CAICT), China's artificial intelligence industry scale has surpassed 900 billion yuan in 2024, representing a year-on-year increase of 24%.
• As of September 2025, the number of AI enterprises in China has exceeded 5,300, accounting for 15% of the global total, forming a complete industrial system. 🚀
• These findings were released on September 23rd at the 2025 Artificial Intelligence Industry and Empowering New Industrialization Conference.

Related Links:

• CAICT: China's AI Industry Scale Exceeds 900 Billion Yuan in 2024, Up 24% Year-on-Year 📈
• Alibaba Announces Collaboration with NVIDIA on Physical AI to Drive Embodied AI Applications
• Alibaba CEO Eddie Wu: A 380 Billion Yuan Three-Year AI Infrastructure Construction Plan is Underway
• Alibaba Cloud Apsara Conference Releases Six Major Models in One Go, Including Qianwen 3-VL and Wanxiang 2.5, Plus New Brand Tongyi Bailing, Covering All Scenarios: Text, Vision, Speech, Video, Code, and Image
• Powered by Google's Strongest Image Generation AI Model: Mixboard Visualizes Your Creativity with Open Canvas + Natural Language Editing ✨

阿里与英伟达联手，加速具身智能开发

简报：

• 在2025阿里云栖大会上，阿里巴巴宣布与英伟达就Physical AI（物理人工智能）展开合作，推动具身智能应用落地。
• 合作内容为阿里云人工智能平台PAI将集成英伟达的Physical AI软件栈，为企业提供数据预处理、仿真、模型训练评估、强化学习等全链路平台服务。
• 此举旨在缩短具身智能、辅助驾驶等应用的开发周期。
• 阿里巴巴CEO吴泳铭在会上透露，公司正积极推进一项为期三年、总投入3800亿元的AI基础设施建设计划。
• 消息宣布后，阿里巴巴港股股价直线拉升，涨幅一度超过9%。

相关链接：

• 阿里巴巴宣布与英伟达开展 Physical AI 合作，推动具身智能应用落地
• 阿里巴巴 CEO 吴泳铭：正在推进三年 3800 亿的 AI 基础设施建设计划

Alibaba and NVIDIA Join Forces to Accelerate Embodied AI Development 🚀

Brief:

• At the 2025 Alibaba Cloud Yunqi Conference, Alibaba announced a partnership with NVIDIA on Physical AI to advance the implementation of embodied AI applications.
• The collaboration involves Alibaba Cloud's AI platform PAI integrating NVIDIA's Physical AI software stack, providing enterprises with full-lifecycle platform services including data preprocessing, simulation, model training and evaluation, and reinforcement learning.
• This initiative aims to shorten the development cycle for applications such as embodied AI and assisted driving.
• Alibaba CEO Eddie Wu (Wu Yongming) revealed at the conference that the company is actively pushing forward a three-year AI infrastructure construction plan with a total investment of 380 billion yuan. 🤖
• Following the announcement, Alibaba's Hong Kong-listed shares saw a sharp increase, with gains exceeding 9% at one point. 📈

Related Links:

• 阿里巴巴宣布与英伟达开展 Physical AI 合作，推动具身智能应用落地
• 阿里巴巴 CEO 吴泳铭：正在推进三年 3800 亿的 AI 基础设施建设计划

通义千问最强模型Qwen3-Max发布，参数超万亿性能领先

简报： 阿里云于9月24日正式发布其迄今规模最大、能力最强的AI大模型通义千问Qwen3-Max，该模型总参数超过一万亿，并在代码生成与智能体工具调用等多个基准测试中达到业界领先水平。

相关链接：

• 阿里云栖大会一口气发布千问 3-VL、万相 2.5 等六大模型 + 通义百聆新品牌，覆盖文本、视觉、语音、视频、代码、图像全场景
• 参数超万亿：阿里发布通义千问最强 AI 大模型 Qwen3-Max 正式版，性能全面领先
• 阿里平頭哥晶片崛起　性能逼近輝達H20獲官方億元大單

Tongyi Qianwen Launches Qwen3-Max, Its Strongest Model Yet, Featuring Over a Trillion Parameters and Leading Performance 🚀

Brief: On September 24, Alibaba Cloud officially unveiled Tongyi Qianwen Qwen3-Max, its largest and most capable AI large model to date. This model boasts over one trillion parameters 🧠 and has achieved industry-leading performance in various benchmark tests, including code generation and agent tool invocation. ✨

Related Links:

• 阿里云栖大会一口气发布千问 3-VL、万相 2.5 等六大模型 + 通义百聆新品牌，覆盖文本、视觉、语音、视频、代码、图像全场景
• 参数超万亿：阿里发布通义千问最强 AI 大模型 Qwen3-Max 正式版，性能全面领先
• 阿里平頭哥晶片崛起　性能逼近輝達H20獲官方億元大單

辉达豪掷千亿美元助OpenAI建AI数据中心，ONO联盟成形

简报：

• 辉达（NVIDIA）与OpenAI宣布达成战略合作，将共同打造一个用电规模达10吉瓦（GW）的下一代AI数据中心。
• 辉达承诺分阶段投入高达1000亿美元（约新台币3兆）的资金，用于支持先进AI模型的开发与运行。
• 该数据中心将配备数百万颗GPU，首个GW规模的系统预计将于2026年下半年上线。
• 此项合作还涉及甲骨文（Oracle），三方形成资金与资源的循环，被外界称为“ONO联盟”闭环。

相关链接：

• 新聞幕後／輝達千億投資成功造浪　ONO聯盟完美金三角誕生？
• 輝達與OpenAI戰略合作　砸3兆打造超級AI運算中心
• H1B簽證費飆！黃仁勳與Altman齊聲「我們需要最聰明的人」

NVIDIA Pledges $100 Billion to Help OpenAI Build AI Data Centers, ONO Alliance Takes Shape 🚀

Brief:

• NVIDIA and OpenAI have announced a strategic partnership to jointly build a next-generation AI data center with a power scale of 10 gigawatts (GW). ⚡️
• NVIDIA has committed to a phased investment of up to $100 billion (approximately NT$3 trillion) to support the development and operation of advanced AI models.
• The data center will be equipped with millions of GPUs, with the first GW-scale system expected to go online in the second half of 2026.
• This collaboration also involves Oracle, forming a tripartite "ONO Alliance" closed loop of funding and resources, as dubbed by external observers. 🤝

Related Links:

• 新聞幕後／輝達千億投資成功造浪　ONO聯盟完美金三角誕生？
• 輝達與OpenAI戰略合作　砸3兆打造超級AI運算中心
• H1B簽證費飆！黃仁勳與Altman齊聲「我們需要最聰明的人」

快手推出可灵AI视频新模型，性能提升且更便宜

简报：

• 快手于2025年9月23日发布其AI视频生成模型的新版本——可灵 2.5 Turbo。
• 新模型在文本理解、动态效果和美学风格上均有提升，能更深入理解复杂指令并生成动态幅度更大的动作。
• 在高品质模式（1080p）下，生成5秒视频的价格相比2.1版本降低了近30%。

相关链接：

• 快手可灵 2.5 Turbo 模型上线，较上代模型便宜近 30%

Kuaishou Launches New Kling AI Video Model: Enhanced Performance and More Affordable 🚀

Brief:

• Kuaishou released a new version of its AI video generation model, Kling 2.5 Turbo, on September 23, 2025.
• The new model boasts improvements in text comprehension, dynamic effects, and aesthetic style. It can now better understand complex instructions and generate actions with significantly greater dynamic range. ✨
• The cost of generating a 5-second video in high-quality mode (1080p) has been reduced by nearly 30% compared to version 2.1. 💰

Related Link:

• Kuaishou Kling 2.5 Turbo Model Launched, Nearly 30% Cheaper Than Previous Generation

联发科发布3nm旗舰芯片天玑9500，主打AI与游戏体验

简报：

• 联发科于9月22日正式发布最新旗舰芯片天玑9500，核心亮点为“5G Agentic AI”。
• 该芯片采用台积电第三代3奈米制程，整合全大核CPU、旗舰级GPU与NPU 990。
• 性能方面，CPU单核性能较上代提升32%，多核提升17%，峰值功耗降低37%；GPU性能提升33%，支持光线追踪技术。
• AI算力较上代提升超过一倍，首次支持在设备端进行4K画质的图片生成及处理大型语言模型。
• 影像方面，支持最高200MP画质输出，并首次支持4K 120FPS防震录影。
• 首批搭载天玑9500的智能手机预计将于2025年第四季度问世。

相关链接：

• 天璣9500正式亮相 聯發科旗艦晶片火力全開震撼登場

MediaTek Unveils 3nm Flagship Chip Dimensity 9500, Focusing on AI and Gaming Experience 🚀

Brief:

• MediaTek officially unveiled its latest flagship chip, the Dimensity 9500, on September 22nd, with "5G Agentic AI" as its core highlight.
• The chip utilizes TSMC's third-generation 3nm process, integrating a full large-core CPU, flagship-grade GPU, and NPU 990.
• In terms of performance, the CPU boasts a 32% increase in single-core performance and a 17% increase in multi-core performance compared to its predecessor, with a 37% reduction in peak power consumption; GPU performance is up by 33% and supports ray tracing technology.
• AI computing power has more than doubled compared to the previous generation, now supporting on-device 4K image generation and large language model processing for the first time. 🧠
• For imaging, it supports up to 200MP output and, for the first time, 4K 120FPS anti-shake video recording. 📸
• The first smartphones powered by the Dimensity 9500 are expected to launch in Q4 2025.

Related Links:

• Dimensity 9500 Officially Debuts: MediaTek's Flagship Chip Unleashes Full Power