单词学习：![[428599e9-9fa8-44d8-9c85-9e6137f8c13f.mp3]]

新闻播报：![[204cd773-ab21-40a0-8e10-737af1faa670.mp3]]

Photoshop 宣布本月内集成谷歌 AI 模型 Nano Banana

简报：

• Adobe 首席设计推广大使 Howard Pinsky 宣布，谷歌的 AI 模型 Nano Banana 将于本月（2025年9月）内登陆 Photoshop。
• 该集成将允许用户把基于提示词的 AI 编辑功能与 Photoshop 现有的选区、图层、蒙版等“非破坏性”工具结合起来使用。
• Nano Banana 是谷歌 Gemini 2.5 Flash Image 的一部分，此前因其能有效避免“恐怖谷效应”且多项表现优于 GPT-4o 而在网络上走红。

相关链接：

• Photoshop 又获一 AI 利器，本月将集成谷歌“最强 P 图模型”Nano Banana

Photoshop Announces Google AI Model Nano Banana Integration This Month 🚀

Brief:

• Adobe Chief Design Advocate Howard Pinsky announced that Google's AI model, Nano Banana ✨, will be integrated into Photoshop this month (September 2025).
• This integration will allow users to combine prompt-based AI editing features with Photoshop's existing "non-destructive" tools such as selections, layers, and masks.
• Nano Banana is part of Google Gemini 2.5 Flash Image and previously gained popularity online for its effectiveness in avoiding the "uncanny valley effect" and outperforming GPT-4o in multiple aspects 🧠.

Related Link:

• Photoshop 又获一 AI 利器，本月将集成谷歌“最强 P 图模型”Nano Banana

Perplexity版权诉讼升级，大英百科与韦氏词典加入起诉

简报：

• 《大英百科全书》与《韦氏词典》于本周三在纽约联邦法院对 AI 搜索公司 Perplexity 提起诉讼，加入了已有的控方阵容。
• 两家公司指控 Perplexity 非法复制其材料侵犯版权，通过 AI 生成的摘要“搭便车”减少版权方收入。
• 诉讼还称，Perplexity 将 AI 产生的幻觉内容归因于这两本工具书，侵犯了其商标权。
• 控方要求法院判决 Perplexity 支付金钱赔偿，并下令停止滥用其内容。

相关链接：

• AI 搜索 Perplexity 深陷版权官司漩涡，大英百科全书和韦氏词典加入控方阵容

Perplexity Copyright Lawsuit Escalates: Britannica and Merriam-Webster Join Suit ⚖️

Brief:

• Encyclopædia Britannica and Merriam-Webster filed a lawsuit against AI search company Perplexity in New York federal court on Wednesday, joining an existing lineup of plaintiffs.
• The two companies accuse Perplexity of illegally copying their materials, infringing copyright, and "free-riding" on AI-generated summaries, thereby reducing revenue for copyright holders.
• The lawsuit also claims that Perplexity attributes AI-generated "hallucinated" content to these two reference works, infringing their trademark rights. 📚
• The plaintiffs are seeking monetary damages from Perplexity and an order to cease the misuse of their content. 🛑

Related Link:

• AI 搜索 Perplexity 深陷版权官司漩涡，大英百科全书和韦氏词典加入控方阵容

ChatGPT 上线开发者模式，Plus/Pro 用户可连接外部系统

简报：

• OpenAI 宣布为 ChatGPT 网页版推出“开发者模式”，该功能目前仅向 Plus 和 Pro 订阅用户开放。
• 用户在设置中启用后，可配置并接入 MCP（Model Context Protocol）客户端，从而直接与外部系统交互。
• 该模式支持更新 CRM 记录、为代码仓库创建拉取请求等高级操作，使应用更贴近真实开发流程。
• 为保障安全，所有可能修改数据的操作都需要用户手动确认，且系统会提示“提示词注入”等潜在风险。

相关链接：

• OpenAI 为 ChatGPT 带来“开发者模式”，Plus / Pro 订阅用户可配置接入 MCP 客户端

ChatGPT Launches Developer Mode, Plus/Pro Users Can Connect External Systems

Brief:

• OpenAI announced the launch of "Developer Mode" for the ChatGPT web version, a feature currently exclusively available to Plus and Pro subscribers. ✨
• Once enabled in the settings, users can configure and connect to an MCP (Model Context Protocol) client, allowing direct interaction with external systems.
• This mode supports advanced operations such as updating CRM records and creating pull requests for code repositories, making applications more aligned with real-world development workflows. 🚀
• To ensure security, all data-modifying operations require manual user confirmation, and the system will alert users to potential risks like "prompt injection." 🔒

Related Links:

• OpenAI brings "Developer Mode" to ChatGPT, Plus / Pro subscribers can configure and connect MCP clients

微软Copilot整合定价，取消销售、服务等附加费

简报：微软宣布，自10月中旬起，此前作为Microsoft 365 Copilot订阅附加功能、每月需额外支付20美元的销售、服务和财务专属Copilot解决方案，将免费向所有订阅用户开放。

相关链接：

• 微软调整 Copilot 定价策略：销售、服务、财务专属功能不再额外收费
• 微软 Copilot 新增音频表达式功能，莎士比亚腔、体育解说都能整

Microsoft Copilot: Pricing Consolidated, Eliminates Sales, Service & Finance Surcharges

Brief: Microsoft has announced a significant update to its Copilot offering! Starting mid-October, the specialized Copilot solutions for sales, service, and finance, previously available as an add-on feature for Microsoft 365 Copilot subscriptions at an additional cost of $20 per month, will now be available free of charge to all subscribers. This move effectively integrates these features into the core Copilot subscription. 🎉💰

Related Links:

• Microsoft Adjusts Copilot Pricing Strategy: Sales, Service, Finance-Specific Features No Longer Extra Charge
• Microsoft Copilot Adds Audio Expression Feature, Can Emulate Shakespearean Accents and Sports Commentary

英伟达市值登顶，创始人因股权稀释财富排名落后

简报：

• Nvidia已成为全球市值最高的公司，但其创始人兼CEO黄仁勋在《彭博亿万富豪榜》上仅排名第十，净资产为1470亿美元。
• 其财富排名与公司市值不符的主要原因是持股比例较低，目前在Nvidia的股份不足4%。
• 黄仁勋的持股因公司长期采用股票作为员工薪酬与激励方式而被稀释，他最初曾持有公司12.8%的股份，若维持至今其身价将超过5000亿美元。
• 相比之下，其他顶级富豪在其公司中持股比例更高，例如马斯克持有特斯拉约13%的股份，伯纳·阿尔诺掌控LVMH约50%的股权。

相关链接：

• 新聞幕後／Nvidia全球市值第一！黃仁勳卻只排富豪榜第十？原因曝光
• 現在轉系還來得及嗎？黃仁勳曝光未來爆賺學科　不是資工而是物理AI

Nvidia Tops Market Cap, Founder's Wealth Ranking Lags Due to Equity Dilution

Brief:

• Nvidia has become the world's most valuable company by market capitalization, yet its founder and CEO, Jensen Huang, ranks only tenth on the Bloomberg Billionaires Index with a net worth of $147 billion. 📈
• The primary reason for this discrepancy between his wealth ranking and the company's market cap is his relatively low ownership stake, currently holding less than 4% of Nvidia's shares. 📉
• Huang's stake has been diluted over time as the company has consistently used stock as a form of employee compensation and incentive. He originally held 12.8% of the company; if he had maintained that stake, his net worth would now exceed $500 billion.
• In comparison, other top billionaires hold significantly larger stakes in their respective companies; for instance, Elon Musk owns approximately 13% of Tesla, and Bernard Arnault controls around 50% of LVMH. 💼

Related Links:

• 新聞幕後／Nvidia全球市值第一！黃仁勳卻只排富豪榜第十？原因曝光
• 現在轉系還來得及嗎？黃仁勳曝光未來爆賺學科　不是資工而是物理AI

阿里云开源通义Qwen3-Next新模型，提升长上下文处理效率

简报：

• 阿里云通义团队于9月12日发布新一代基础模型架构Qwen3-Next，并开源了Qwen3-Next-80B-A3B系列模型。
• 新架构通过改进混合注意力机制和采用高稀疏度MoE结构，旨在提升模型在长上下文和大规模参数下的训练与推理效率。
• 开源的Qwen3-Next-80B-A3B模型拥有800亿参数，但推理时仅激活30亿，其训练成本不到Qwen3-32B模型的十分之一，长上下文推理吞吐量则提升十倍以上。
• 该模型原生支持262K上下文，并可外推至约101万tokens。
• 模型权重已在Hugging Face上以Apache-2.0许可发布，并可通过多种框架部署。

相关链接：

• 阿里云发布通义 Qwen3-Next 基础模型架构并开源 80B-A3B 系列：改进混合注意力机制、高稀疏度 MoE 结构

Alibaba Cloud Open-Sources New Tongyi Qwen3-Next Model, Enhancing Long-Context Processing Efficiency ✨

Brief:

• The Alibaba Cloud Tongyi team unveiled the new generation foundational model architecture, Qwen3-Next, on September 12, and open-sourced the Qwen3-Next-80B-A3B series models.
• The new architecture aims to enhance the model's training and inference efficiency for long-context and large-scale parameters through an improved hybrid attention mechanism and the adoption of a high-sparsity Mixture-of-Experts (MoE) structure.
• The open-sourced Qwen3-Next-80B-A3B model boasts 80 billion parameters but activates only 3 billion during inference. Its training cost is less than one-tenth of the Qwen3-32B model, while long-context inference throughput is boosted by over tenfold. 🚀
• The model natively supports 262K context and can be extrapolated to approximately 1.01 million tokens.
• Model weights have been released on Hugging Face under the Apache-2.0 license and can be deployed via multiple frameworks. ⚙️

Related Links:

• Alibaba Cloud Releases Tongyi Qwen3-Next Foundational Model Architecture and Open-Sources 80B-A3B Series: Improved Hybrid Attention Mechanism, High-Sparsity MoE Structure