我们在八月份发布的最新人工智能资讯
内容来源:https://blog.google/technology/ai/google-ai-updates-august-2025/
内容总结:
谷歌八月人工智能动态盘点:多领域创新成果集中亮相
在持续二十余年对机器学习与人工智能的研发投入基础上,谷歌于八月推出多项突破性AI技术进展。本月发布的成果涵盖搜索体验升级、硬件创新、图像处理革新及教育工具普惠等多个维度,全面体现其"以AI技术提升人类生活质量"的核心目标。
搜索服务方面,AI搜索模式(AI Mode)已扩展至全球180多个国家和地区,新增个性化餐饮推荐功能与协作分享按钮,使搜索引擎具备更强大的代理交互能力。同期,Gemini应用面向Ultra订阅用户推出"深度思考"(Deep Think)高阶推理功能,其底层模型更在国际数学奥林匹克竞赛中达到金牌标准。
硬件生态领域,谷歌在年度发布会上推出搭载Tensor G5芯片的Pixel 10系列手机,全线产品集成Gemini Nano端侧AI能力,支持魔法提示、语音图像编辑等智能功能。配套的NotebookLM工具可生成定制化音视频产品解读方案。
图像处理技术获得重大突破——DeepMind研发的"Nano Banana"图像编辑模型正式集成至Gemini应用,支持人像一致性编辑与多风格融合创作。同时,全球首个实时交互通用世界模型Genie 3问世,该模型通过文本提示即可生成复杂交互环境,为机器人训练与AGI研究提供新路径。
教育支持方面,谷歌向美国、日韩、印尼、巴西等多国大学生免费提供AI Pro计划年度订阅,包含Gemini 2.5 Pro、Veo 3等高级工具的使用权限。此外,异步编程助手Jules结束测试全面开放,Translate应用新增AI实时翻译与语言练习功能。
内部应用层面,谷歌同步发布了14项员工AI工作效率提升方案,涵盖创意激发、代码生成等实际应用场景,展现人工智能技术在企业运营中的深度整合。
中文翻译:
我们于八月发布的最新人工智能资讯
二十余年来,我们持续投入机器学习与人工智能领域的研究、工具及基础设施建设,致力于打造能普惠大众生活的产品。谷歌各团队正积极探索人工智能在医疗健康、危机应对、教育等广泛领域的应用价值。为及时同步我们的进展,我们将定期汇总谷歌最新人工智能动态。
以下是八月部分重要发布的回顾:
这个八月可谓硕果累累。我们不仅拓展了搜索中的AI模式,推出了搭载尖端设备端人工智能技术的Pixel新品系列,还带来了诸多亮点——包括Gemini应用中的全新图像编辑功能(Nano Banana),以及谷歌DeepMind首款实时交互通用世界模型Genie 3。本月我们发布了多个人工智能相关进展,涵盖助力濒危物种保护等技术应用,所有努力都聚焦于同一使命:通过人工智能产品与功能提升全球用户的生活品质。
我们扩展了搜索中的AI模式并新增代理功能。AI模式先进的代理与个性化能力意味着谷歌能为您代劳更多事务。目前在美国实验室功能中启用AI模式实验的用户,可体验基于个人偏好定制的餐饮推荐结果。通过新增的"分享"按钮,您还能与好友协作并共享AI模式链接。此外,我们已将AI模式扩展至全球180多个国家和地区(英文版)。
我们为Google AI Ultra订阅用户推出了Gemini应用中的"深度思考"功能。这项强大的复杂推理与问题解决工具现已面向Ultra用户开放。该模型的变体更在今年国际数学奥林匹克竞赛(IMO)中达到金牌标准!
我们在2025年谷歌制造大会上发布了搭载最强人工智能的Pixel硬件。年度硬件盛会见证了迄今最智能手机系列的诞生:搭载谷歌Tensor G5芯片的AI驱动型Pixel 10、Pixel 10 Pro、Pixel 10 Pro XL及Pixel 10 Pro Fold。新款Pixel内置Gemini Nano,支持魔法提示词、文本语音指令修图、Gemini Live智能视觉助手等创新功能。
若觉链接过多,只需点击此处:通过谷歌制造笔记本功能,NotebookLM可通过音视频概述解读发布会内容,还能即时生成定制化操作指南。
我们在Gemini应用中推出了新型图像模型。谷歌DeepMind研发的"Nano Banana"图像生成编辑模型已是全球评分最高的修图工具。现用户可通过Gemini应用编辑旧照片或创作新图像,在更换服饰、融合照片、混搭风格时保持人物乃至宠物的一致性特征。
我们向全球大学生免费开放最先进的人工智能学习工具。随着开学季来临,我们为美国、日本、印尼、韩国、巴西等国高校生提供Google AI Pro计划一年免费订阅服务。学生可畅享Gemini 2.5 Pro、深度研究、NotebookLM、Veo 3及引导式学习等工具,助力培养批判性思维、深化知识理解与提升学习效能。
我们分享了14种谷歌员工提升工作效率的AI应用技巧。通过调研员工如何运用AI激发创意、创作内容、完成代码生成与数据分析等任务,我们将实践心得转化为实用建议面向公众发布。
我们向全民开放异步编程助手Jules。Jules被设计为能集成代码库、理解用户意图、制定计划并执行任务的自主代理工具。经过Beta测试阶段的界面优化与代码修正后,现正式开放支持所有用户的编程需求。
我们在谷歌翻译中推出AI实时翻译与语言学习工具。基于先进AI模型的全新实时翻译功能支持70余种语言的即时交流。此外,新增的语言练习功能将助您更高效达成学习目标。
我们发布了通用世界模型Genie 3。这款突破性模型仅通过文本提示即可生成多样化的交互环境。世界模型作为通向通用人工智能(AGI)的关键基石,能为AI智能体提供无限丰富的模拟训练环境。Genie 3不仅可为机器人及自动驾驶系统等提供训练场景,更开创了评估智能体性能与探索其缺陷的新范式。您可通过Release Notes播客收听谷歌DeepMind首席执行官Demis Hassabis对Genie 3的深度解读。
英文来源:
The latest AI news we announced in August
For more than 20 years, we’ve invested in machine learning and AI research, tools and infrastructure to build products that make everyday life better for more people. Teams across Google are working on ways to unlock AI’s benefits in fields as wide-ranging as healthcare, crisis response and education. To keep you posted on our progress, we're doing a regular roundup of Google's most recent AI news.
Here’s a look back at some of our AI announcements from August.
August was bananas. We had an AI Mode in Search expansion, a new line of Pixel hardware that includes our most advanced on-device AI and plenty of other news to round out the month — including a new image editing release in the Gemini app (Nano Banana) and Google DeepMind’s first real-time interactive general-purpose world AI model, Genie 3. We’ve made bunches of AI news this month, including AI to help endangered species, and it’s all focused on our mission of delivering AI products and features in a way that can improve the lives of as many people as possible.
We expanded AI Mode in Search and added new agentic features. AI Mode’s new advanced agentic and personalized capabilities mean Google can do more on your behalf and help you get things done. Now, if you’ve opted into the AI Mode experiment in Labs in the U.S., you will see results tailored to your personal preferences, beginning with dining recommendations. You can even collaborate and share your AI Mode links with friends using the new “Share” button. We’ve also expanded AI Mode to more than 180 new countries and territories in English.
We rolled out Deep Think in the Gemini app for Google AI Ultra subscribers. Deep Think can be an incredibly powerful tool for complex reasoning and problem solving, and we made it available to Google AI Ultra subscribers. A variation of this same model won the gold-medal standard at this year’s International Mathematical Olympiad (IMO)!
We rolled out the best Google AI in Pixel hardware at Made by Google 2025. Our annual hardware event featured the release of our most helpful phones yet, the AI-powered Pixel 10, Pixel 10 Pro, Pixel 10 Pro XL and Pixel 10 Pro Fold, all featuring the Google Tensor G5 chip. Our latest Pixels come with Gemini Nano to power features like Magic Cue, photo editing with text and voice prompts, Gemini Live for a more helpful, natural and visual assistant and more.
And if that’s too many links to click through, click just this one: With our Made by Google notebook, you can have NotebookLM explain our Made by Google announcements with an Audio or Video Overview, and even get how-to articles custom-made in seconds.
We unveiled a new image model in the Gemini app. Google DeepMind’s image generation and editing model “Nano Banana” is already the top-rated image editing model in the world. Now, it’s in the Gemini app, where you can edit your old photos or create new images. You can maintain a consistent likeness when editing photos of people or even pets, all while switching up outfits, blending photos and mixing up styles.
We made our most advanced AI learning tools available to college students for free. As back to school season kicked off, we made a free one-year subscription to Google’s AI Pro plan available to college students in the U.S. as well as Japan, Indonesia, Korea and Brazil — and more. Students get expanded access to tools Gemini 2.5 Pro, Deep Research, NotebookLM, Veo 3 and Guided Learning to help foster critical thinking, deepen understanding and improve studying.
We shared 14 ways Googlers use AI to work smarter. After asking Googlers how they use AI every day to spark creativity, create content, complete tasks like code generation and data analysis and more, we turned what we found into tips for you.
We made our asynchronous coding agent, Jules, available for everyone. Jules was designed to be an autonomous agent that integrates with your repositories, understands your intent, makes a plan and gets to work. After refining Jules in beta with a polished user interface and code fixes, we made it available to support everyone with coding tasks.
We introduced AI-powered live translation and language learning tools in Google Translate. Using advanced AI models, the new live capabilities in Translate make it easier to communicate in real time in more than 70 languages. Plus, to help you reach your learning goals, we added a new language practice feature.
We announced Genie 3, a general purpose world AI model. Genie 3 is Google DeepMind’s groundbreaking general purpose world model that can generate diverse, interactive environments from a single text prompt. World models are also a key stepping stone on the path to artificial general intelligence (AGI), since they make it possible to train AI agents in an unlimited curriculum of rich simulation environments. Not only can Genie 3 provide environments to train agents like robots and autonomous systems, but it could also offer new ways to evaluate agents’ performance, and explore their weaknesses. You can listen to Google DeepMind CEO Demis Hassabis talk about Genie 3 on the Release Notes podcast.