![[93787673-77a7-46da-bc78-14442651c81e.mp3]]

ChatGPT Projects重大升级：引入深度研究与语音交互等多项新功能

简报：

• 2025年6月12日，OpenAI在“12 Days of OpenAI”活动中对ChatGPT Projects功能进行重大升级，将其由原本的对话和资料整理工具升级为智能工作空间。
• 新增的深度研究功能允许聊天项目结合网页搜索与项目内部数据，实现更精准的信息检索与内容整合。
• 集成语音模式，用户可通过语音与项目直接交互，提升移动场景和实时协作的便捷性。
• 优化记忆功能，ChatGPT可更智能地引用历史对话和上传内容，提升项目执行效率和连续性。
• 移动端支持增强，支持文件上传、模型切换、实时视频与屏幕共享，进一步拓展了移动办公和协作场景。
• 升级后的Projects适用于个人任务管理、团队协作、教育研究和企业应用，并新增与企业平台（如HubSpot）的集成能力，提升商业场景实用性。

相关链接：

• OpenAI官方帮助页面 - 使用 ChatGPT Projects
• 功能解读报道
• Venturebeat：OpenAI launches ChatGPT Projects, letting you organize files, chats in groups

ChatGPT Projects Undergoes Major Upgrade: Introducing Deep Research, Voice Interaction, and More New Features

Briefing:

• On June 12, 2025, OpenAI significantly upgraded the ChatGPT Projects feature during its "12 Days of OpenAI" event, transforming it from a conversational and data organization tool into an intelligent workspace. 🚀
• The newly added Deep Research feature enables chat projects to combine web search with internal project data, achieving more precise information retrieval and content integration.
• Integrated Voice Mode allows users to interact directly with projects via voice, enhancing convenience for mobile scenarios and real-time collaboration. 🗣️
• Optimized Memory function enables ChatGPT to more intelligently reference historical conversations and uploaded content, improving project execution efficiency and continuity.
• Enhanced Mobile Support now includes file uploads, model switching, real-time video, and screen sharing, further expanding mobile office and collaboration scenarios.
• The upgraded Projects are suitable for personal task management, team collaboration, educational research, and enterprise applications. New integration capabilities with enterprise platforms (e.g., HubSpot) have also been added, enhancing practical utility in business scenarios. ✨

Related Links:

• OpenAI Official Help Page - Using ChatGPT Projects
• Feature Interpretation Report (Chinese)
• Venturebeat: OpenAI launches ChatGPT Projects, letting you organize files, chats in groups

谷歌Imagen4模型集成Gemini平台，用户免费体验专业级AI图像生成

简报：

• 谷歌最新图像生成模型Imagen4已正式集成至Gemini平台，用户可免费体验专业级AI图像创作功能；
• Imagen4在高分辨率细节、智能英文文本渲染和多样化色彩表现上取得显著突破，可应对海报、广告等多种场景的设计需求；
• 用户可在Gemini聊天界面通过文本提示生成和实时优化图像，极大简化创作流程；
• 部分用户反馈非英文文本渲染尚有提升空间，但整体英文支持已趋成熟；
• 这一集成不仅丰富了Gemini平台的AI生态，也为内容创作者与企业团队降低了图像创作门槛，提高了效率。

相关链接：

• Gemini 集成 Imagen4：免费解锁专业级图像创作

Google's Imagen4 Model Integrated into Gemini Platform, Users Can Experience Professional AI Image Generation for Free

Brief:

• Google's latest image generation model, Imagen4, has been officially integrated into the Gemini platform, allowing users to experience professional-grade AI image creation features for free. ✨
• Imagen4 has achieved significant breakthroughs in high-resolution detail, intelligent English text rendering, and diverse color expression, capable of meeting design demands for various scenarios like posters and advertisements.
• Users can generate and optimize images in real-time through text prompts within the Gemini chat interface, greatly simplifying the creative process. 🚀
• Some users have reported that non-English text rendering still has room for improvement, but overall English support is nearing maturity.
• This integration not only enriches Gemini's AI ecosystem but also lowers the barrier for image creation for content creators and enterprise teams, significantly boosting efficiency. 💡

Related links:

• Gemini Integrates Imagen4: Unlock Professional Image Creation for Free

谷歌新AI模型提升气候预测精度至10公里

简报：

• 谷歌研究人员创新性地结合传统物理气候建模与生成式AI，推出“动态生成下采样法”，大幅提升气候预测的分辨率至约10公里。
• 该方法通过两步流程，先用物理模型下采样到中等分辨率，再由R2D2生成AI模型还原高分辨率细节，显著增强区域气候风险评估的准确性和效率。
• 实验显示，新模型在温度、湿度和风速等预测上较传统方法误差降低40%以上，捕捉复杂天气模式能力更强，同时计算成本相比传统高分辨率模拟下降高达85%，支持多领域的精细化应用。

相关链接：

• 谷歌 AI 助力气候预测：打破传统模型的局限，精确到 10 公里！

Google's New AI Model Boosts Climate Prediction Accuracy to 10 Kilometers

Briefing:

• Google researchers have innovatively combined traditional physical climate modeling with generative AI, introducing the "Dynamic Generative Downsampling Method" to significantly enhance climate prediction resolution to approximately 10 kilometers. 🌍
• This method employs a two-step process: first, downsampling to medium resolution using physical models, then having the R2D2 generative AI model restore high-resolution details. This significantly boosts the accuracy and efficiency of regional climate risk assessments. 📈
• Experiments show that the new model reduces errors in temperature, humidity, and wind speed predictions by over 40% compared to traditional methods, demonstrating a stronger ability to capture complex weather patterns. Simultaneously, computational costs are reduced by up to 85% compared to traditional high-resolution simulations, supporting fine-grained applications across multiple domains. 💡

Related Links:

• 谷歌 AI 助力气候预测：打破传统模型的局限，精确到 10 公里！

新型脑机接口助渐冻症患者实时合成语音与旋律

简报：

• 美国加州大学戴维斯分校团队研发的新型脑机接口，在一名肌萎缩侧索硬化症（渐冻症）患者身上试验成功，实现了患者通过电脑几乎无延迟地重新流畅说话和哼唱旋律。
• 该系统通过植入微电极实时捕捉大脑发声区域的神经信号，并结合AI解码模型，将这些信号即时转化为类似患者本人声音的语音输出，甚至可以控制语调、复现语气词和合成虚构词语，使交流更自然。
• 技术还可根据大脑活动识别患者试图唱歌的音高，合成旋律，为丧失语言能力的人群带来新的沟通希望。
• 本项技术目前仅在一名患者身上测试，研究团队计划将研究拓展至更多丧失语言能力的群体，以验证其普适性。

相关链接：

• IT之家报道
• 36氪快讯

Novel Brain-Computer Interface Enables Real-time Speech and Melody Synthesis for ALS Patients

Brief:

• A novel brain-computer interface (BCI) developed by a team at the University of California, Davis, USA, has been successfully tested on an Amyotrophic Lateral Sclerosis (ALS) patient, enabling the patient to regain fluent speech and hum melodies through a computer with virtually no delay. 🧠
• The system captures neural signals from the brain's speech production areas in real-time via implanted microelectrodes. Combined with an AI decoding model, it instantly converts these signals into speech output resembling the patient's own voice. It can even control intonation, reproduce interjections, and synthesize fictional words, making communication more natural. 🗣️
• The technology can also identify the pitch of melodies the patient attempts to sing based on brain activity, synthesizing them, bringing new hope for communication to individuals who have lost their ability to speak. 🎶
• Currently, this technology has only been tested on one patient. The research team plans to expand the study to more individuals with lost speech ability to verify its universality.

Related Links:

• ITHome Report (Chinese)
• 36Kr Newsflash (Chinese)