![[a58edfeb-21f6-4551-9ee3-449c7ffbd1bd.mp3]]

Shopify推出AI工具Shopify Magic，助力商家快速搭建在线商店

简报：

• Shopify推出名为Shopify Magic的AI工具，可通过关键词自动生成产品描述、邮件主题和商店标题
• 该工具基于大型语言模型(LLM)，能分析关键词并模仿人类语言生成专业文案
• 商家可选择不同语气风格，确保品牌信息一致性
• Shopify Magic是该公司首次进军生成式AI领域，被视为与亚马逊竞争的最新举措
• 该工具目前主要支持英语，未来可能扩展更多语言功能

相关链接：

• 用人工智能帮助卖家编写产品描述，Shopify推出生成式AI产品

• Shopify推出Shopify Magic，提升撰写效率

• Shopify Magic助力撰写效率，未来AI工具趋势与跨境卖家影响

Shopify Launches AI Tool Shopify Magic to Help Merchants Build Online Stores Quickly

Newsletter:

• Shopify introduces an AI tool named Shopify Magic, which automatically generates product descriptions, email subjects, and store titles based on keywords.
• The tool is powered by large language models (LLM), capable of analyzing keywords and mimicking human language to create professional copy.
• Merchants can select different tone styles to ensure consistency in brand messaging.
• Shopify Magic marks the company’s first venture into generative AI, seen as a latest move to compete with Amazon. 💡
• Currently, the tool primarily supports English, with potential expansion to more languages in the future. 🌍

Related Links:

• Using AI to Help Sellers Write Product Descriptions, Shopify Launches Generative AI Product
• Shopify Launches Shopify Magic, Boosting Writing Efficiency
• Shopify Magic Enhances Writing Efficiency, Future AI Tool Trends and Impact on Cross-Border Sellers 🚀

苹果将在WWDC25开放AI模型权限，允许开发者集成设备端轻量模型

简报：

• 苹果计划在6月9日WWDC25上宣布向第三方开发者开放AI模型权限，初期将聚焦设备端运行的轻量级模型（约30亿参数）
• 开放模型包括本地部署版本（iPhone 15 Pro上可达30token/s生成速度）和云端版本（性能对标GPT-4-Turbo）
• 开发者可将Apple Intelligence技术集成到应用中，如改进Siri交互、添加AI图像生成或文本编辑功能
• 此举是苹果应对AI领域竞争的战略调整，目前App Store有600万开发者，预计10%将开发AI应用
• 隐私保护方面，云端数据传输需用户授权并采用加密技术，本地模型处理可避免数据上传

相关链接：

• 苹果坐不住了！将向开发者开放AI大模型
• 苹果准备放出AI底牌：向开发者开放大模型权限
• 终于要来了！苹果将在六月份全面开放AI功能

Apple to Open AI Model Access at WWDC25, Allowing Developers to Integrate Lightweight On-Device Models

Briefing:

• Apple plans to announce the opening of AI model access to third-party developers at WWDC25 on June 9, initially focusing on lightweight models running on-device (approximately 3 billion parameters) 🤖
• The opened models include an on-device version (capable of up to 30 tokens/s generation speed on iPhone 15 Pro) and a cloud-based version (performance comparable to GPT-4-Turbo)
• Developers will be able to integrate Apple Intelligence technology into their apps, such as enhancing Siri interactions, adding AI image generation, or text editing features 🎨
• This move marks a strategic shift by Apple to address competition in the AI field, with 6 million developers currently on the App Store, of whom an estimated 10% are expected to develop AI applications
• For privacy protection, cloud data transmission requires user consent and employs encryption technology, while on-device model processing avoids data uploads 🔒

Related Links:

• Apple Can't Sit Still! Set to Open AI Models to Developers
• Apple Prepares to Reveal AI Trump Card: Opening Large Model Access to Developers
• It's Finally Coming! Apple to Fully Open AI Features in June

OpenAI以65亿美元收购苹果前设计总监Jony Ive的AI硬件公司io

简报：

• OpenAI宣布以65亿美元全资收购前苹果首席设计师Jony Ive创立的AI设备公司io，这是OpenAI迄今为止最大规模的收购
• 交易完成后，Jony Ive将加入OpenAI并负责整体创意与设计工作，其设计工作室LoveFrom将保持独立运营
• io团队约55名员工将整体并入OpenAI，其中包括多位前苹果核心设计师，曾参与iPhone等标志性产品开发
• 此次收购将推动OpenAI进军硬件领域，计划开发"跳脱屏幕"的AI交互设备，首款产品预计2026年发布
• OpenAI此前已持有io公司23%股份，此次实际支付约50亿美元收购剩余股份

相关链接：

• 突发！OpenAI以64亿美元，收购iPhone设计师公司io
• 重磅！OpenAI联合苹果传奇设计大佬Jony Ive官宣新公司「io」
• 苹果传奇人物Jony Ive加入 OpenAI 领导设计工作

OpenAI Acquires Apple’s Former Design Chief Jony Ive’s AI Hardware Company io for $6.5 Billion

Brief:

• OpenAI announces the full acquisition of io, an AI device company founded by former Apple Chief Design Officer Jony Ive, for $6.5 billion, marking OpenAI's largest acquisition to date. 💰
• Following the deal, Jony Ive will join OpenAI to oversee overall creative and design efforts, while his design studio, LoveFrom, will continue to operate independently.
• io’s team of approximately 55 employees will fully integrate into OpenAI, including several former Apple core designers who contributed to iconic products like the iPhone. 📱
• This acquisition will propel OpenAI into the hardware sector, with plans to develop "screen-free" AI interaction devices, targeting the release of the first product by 2026.
• OpenAI previously held a 23% stake in io, with this transaction involving an actual payment of about $5 billion for the remaining shares. 🚀

Related Links:

• Breaking! OpenAI Acquires iPhone Designer’s Company io for $6.4 Billion
• Major News! OpenAI Partners with Apple Legend Jony Ive to Announce New Company ‘io’
• Apple Icon Jony Ive Joins OpenAI to Lead Design Efforts

谷歌搜索AI模式在美国正式上线，集成Gemini 2.5和深度搜索功能

简报：

• 谷歌在Google I/O 2025大会上宣布搜索AI模式正式上线，即日起逐步向美国用户推出
• 新功能整合Gemini 2.5模型，支持处理更长、更复杂的查询（用户提问长度增加2-3倍）
• 深度搜索功能将于今年夏季推出，可更深入探索网页内容
• 新增Search Live功能（夏末推出），支持通过手机摄像头实时搜索
• AI模式提供个性化服务，可学习用户偏好（如餐厅座位选择）并整合个人上下文信息
• 数据显示：使用AI概览的查询量增长10%，Google Lens查询量同比增长65%

相关链接：

• 谷歌搜索推出全新AI模式 含Gemini 2.5、深度搜索
• 谷歌搜索全新AI模式亮相：Gemini 2.5、个性化回答、深度搜索等
• 谷歌I/O 2025大会：更激进的AI搜索变革、更贵的订阅策略

Google Search AI Mode Officially Launches in the US, Integrating Gemini 2.5 and Deep Search Features

Brief:

• Google announced the official launch of Search AI Mode at the Google I/O 2025 conference, rolling out gradually to US users starting today 🌟
• The new feature integrates the Gemini 2.5 model, supporting longer and more complex queries (user question length increased by 2-3 times)
• Deep Search functionality will be introduced this summer, enabling deeper exploration of web content 🔍
• New Search Live feature (launching late summer) will allow real-time searches via smartphone cameras 📸
• AI Mode offers personalized services, learning user preferences (e.g., restaurant seating choices) and integrating personal context information
• Data shows: Queries using AI Overview have increased by 10%, and Google Lens queries have surged by 65%

Related Links:

• Google Search Launches New AI Mode with Gemini 2.5 and Deep Search
• Google Search Unveils New AI Mode: Gemini 2.5, Personalized Answers, Deep Search, and More
• Google I/O 2025 Conference: Bolder AI Search Innovations and Pricier Subscription Strategies

英伟达发布Cosmos-Reason1系列模型，提升AI物理推理能力

简报：

• 英伟达推出Cosmos-Reason1系列模型，包括7B和56B两个版本，专注于提升AI在物理常识和具身推理方面的能力
• 该模型通过Physical AI监督微调（SFT）和强化学习（RL）两阶段训练，采用双本体系统划分物理常识为空间、时间和基础物理三大类
• 模型架构采用仅解码器的大型语言模型（LLM），结合视觉编码器处理视频数据，实现文本与视觉数据的同步推理
• 训练数据包含约400万条标注视频-文本对，涵盖动作描述和复杂推理任务
• 该技术有望应用于机器人、自动驾驶车辆等领域，解决AI在物理世界中理解重力、空间关系等难题

相关链接：

• 英伟达推出 Cosmos-Reason1 系列模型，让 AI 理解物理常识
• 英伟达Cosmos-Reason1：AI物理推理的新突破，让机器人更聪明！
• 推理延展到真实物理世界，英伟达Cosmos-Reason1：8B具身推理表现超过OpenAI ο1

NVIDIA Releases Cosmos-Reason1 Series Models, Enhancing AI Physical Reasoning Capabilities

Brief:

• NVIDIA introduces the Cosmos-Reason1 series models, available in 7B and 56B versions, focusing on improving AI capabilities in physical commonsense and embodied reasoning. 🚀
• The model undergoes a two-stage training process with Physical AI Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), utilizing a dual-ontology system that categorizes physical commonsense into spatial, temporal, and fundamental physics domains.
• The model architecture employs a decoder-only Large Language Model (LLM) paired with a visual encoder to process video data, enabling synchronized reasoning of text and visual inputs.
• Training data comprises approximately 4 million annotated video-text pairs, covering action descriptions and complex reasoning tasks.
• This technology holds potential for applications in robotics and autonomous vehicles, addressing AI challenges in understanding gravity, spatial relationships, and more in the physical world. 🤖🚗

Related Links:

• NVIDIA Launches Cosmos-Reason1 Series Models, Enabling AI to Understand Physical Commonsense
• NVIDIA Cosmos-Reason1: A New Breakthrough in AI Physical Reasoning, Making Robots Smarter!
• Reasoning Extends to the Real Physical World, NVIDIA Cosmos-Reason1: 8B Embodied Reasoning Surpasses OpenAI o1

智元机器人灵犀X2今日开售，三版本售价10万起

简报：

• 智元机器人灵犀X2于5月22日正式开售，提供交互版、探索版和旗舰版三个版本
• 售价区间为10万至30万元不等，具体价格根据版本配置有所不同
• 灵犀X2身高1.3米，具备高自由度运动能力和多模态交互方式
• 适用于文娱商演、展厅导览、科研教育、情感陪伴等多种应用场景
• 用户可通过智元机器人官网进行预订

相关链接：

• 智元机器人灵犀X2即将上市，三款机型售价十万起，探索未来生活新可能！
• 智元：5月22日灵犀X2机器人官网开售

Zhiyuan Robot Lingxi X2 Goes on Sale Today, Starting at 100,000 RMB for Three Versions

Brief:

• Zhiyuan Robot Lingxi X2 officially goes on sale on May 22, available in Interactive, Explorer, and Flagship versions
• Price range varies from 100,000 to 300,000 RMB, depending on the version and configuration
• Lingxi X2 stands at 1.3 meters tall, featuring high-degree-of-freedom mobility and multimodal interaction capabilities
• Suitable for various application scenarios, including entertainment performances, exhibition hall guiding, scientific research, education, and emotional companionship 🤖
• Users can place pre-orders through the official Zhiyuan Robot website 🌐

Related Links:

• Zhiyuan Robot Lingxi X2 Set to Launch, Three Models Starting at 100,000 RMB, Exploring New Possibilities for Future Living!
• Zhiyuan: Lingxi X2 Robot Officially on Sale on May 22 via Official Website 😊

摩尔线程国产GPU算力专区上线AutoDL平台，面向AI开发者开放

简报：

• 摩尔线程与AutoDL平台合作推出面向个人开发者的"摩尔线程专区"，首次将国产GPU算力开放至AI开发一线
• 专区依托摩尔线程夸娥（KUAE）智算集群的全功能GPU算力支持，为开发者提供高性价比的国产算力解决方案
• 专区采用"社区驱动生态共建"模式，首批上线模型由开发者自主适配，覆盖计算机视觉、自然语言处理等热门领域
• 开发者可通过平台一键部署国产算力环境，支持基础镜像与社区镜像灵活调用
• 摩尔线程联合AutoDL推出限时优惠活动，为科研人员、高校学生及个人开发者提供算力折扣

相关链接：

• 摩尔线程算力专区正式上线AutoDL平台，专属优惠同步启动
• 摩尔线程与青云科技达成战略合作，共建国产AI算力新生态

Moore Threads Domestic GPU Computing Zone Launches on AutoDL Platform, Open to AI Developers

Newsletter:

• Moore Threads collaborates with the AutoDL platform to launch the "Moore Threads Zone" for individual developers, marking the first time domestic GPU computing power is made available to frontline AI development 🚀
• The zone leverages the full-featured GPU computing power of Moore Threads’ KUAE intelligent computing cluster, offering developers cost-effective domestic computing solutions 💻
• Operating on a "community-driven ecosystem co-building" model, the initial batch of models is independently adapted by developers, covering popular fields like computer vision and natural language processing 📊
• Developers can deploy domestic computing environments with one click via the platform, supporting flexible use of base images and community images
• Moore Threads, in partnership with AutoDL, introduces limited-time discount campaigns, providing computing power discounts for researchers, university students, and individual developers 🎉

Related Links:

• Moore Threads Computing Zone Officially Launches on AutoDL Platform with Exclusive Discounts
• Moore Threads and Qingyun Technology Reach Strategic Cooperation to Build a New Domestic AI Computing Ecosystem

谷歌开源MedGemma医疗AI模型，支持图像分类与文本分析

简报：

• 谷歌在2025年I/O开发者大会上推出开源医疗AI模型MedGemma，基于Gemma 3架构
• 提供两种配置：4B参数多模态模型（医疗图像分类与解读）和27B参数纯文本模型（临床文本理解）
• 4B模型采用SigLIP图像编码器，预训练数据涵盖胸部X光、皮肤病学、眼科及组织病理学图像
• 27B模型专注于患者分诊、临床决策辅助等深度文本分析任务
• 支持本地实验和Google Cloud Vertex AI部署，提供Colab笔记本等工具辅助微调

相关链接：

• 谷歌开源MedGemma AI模型，医疗图像与文本分析神器
• 谷歌开源MedGemma医疗AI，图像文本分析助力医疗效率大提升
• 谷歌MedGemma开源引领医疗AI技术革新突破

Google Open-Sources MedGemma Medical AI Model, Supporting Image Classification and Text Analysis

Brief:

• Google unveiled the open-source medical AI model MedGemma at the 2025 I/O Developer Conference, built on the Gemma 3 architecture. 🌐
• Offers two configurations: a 4B parameter multimodal model (for medical image classification and interpretation) and a 27B parameter text-only model (for clinical text comprehension).
• The 4B model utilizes the SigLIP image encoder, with pre-training data covering chest X-rays, dermatology, ophthalmology, and histopathology images. 🩻
• The 27B model focuses on deep text analysis tasks such as patient triage and clinical decision support.
• Supports local experimentation and deployment on Google Cloud Vertex AI, with tools like Colab notebooks provided for fine-tuning. 💻

Related Links:

• Google Open-Sources MedGemma AI Model, a Powerhouse for Medical Image and Text Analysis
• Google’s Open-Source MedGemma Medical AI Boosts Healthcare Efficiency with Image and Text Analysis
• Google’s MedGemma Open-Source Model Leads Breakthroughs in Medical AI Technology