![[9db833cf-7d09-4962-ad7d-c6c00910163e.mp3]]

OpenAI与英国政府签署合作推动AI技术在公共服务领域落地

简报：

• OpenAI首席执行官萨姆·阿尔特曼与英国科学与技术大臣彼得·凯尔于本周签署合作备忘录，将探索先进人工智能在英国政府司法、安全、教育等领域的应用，提升公务员和市民高效使用公共服务的能力。
• 此次合作旨在共同开发针对司法、国防安全、教育技术等方面的AI解决方案，并扩大公众对AI技术的了解和参与。
• 该协议是在Google与英国政府签署类似合作协议之后达成的，目前OpenAI在英国已有超100名员工，计划进一步扩展在英业务。
• 英国公众对AI技术发展呈现期待与担忧并存的态度。

相关链接：

• OpenAI与英国政府签署合作协议
• OpenAI与英国签战略协议 探索AI公共服务应用
• OpenAI加码英国，与政府展开AI合作

OpenAI and UK Government Sign Partnership to Advance AI in Public Services 🤝

Brief:

• OpenAI CEO Sam Altman and UK Secretary of State for Science, Innovation and Technology Peter Kyle signed a Memorandum of Understanding this week. The partnership will explore the application of advanced AI in areas such as justice, security, and education within the UK government, enhancing the ability of civil servants and citizens to efficiently use public services. 💡
• This collaboration aims to jointly develop AI solutions for justice, national security, and educational technology, while also expanding public understanding and engagement with AI technologies.
• This agreement follows a similar partnership signed between Google and the UK government. OpenAI currently has over 100 employees in the UK and plans to further expand its operations there.
• The UK public holds a mixed attitude towards the development of AI technology, with both anticipation and concerns. 🤔

Related Links:

• OpenAI Signs Partnership with UK Government
• OpenAI Signs Strategic Agreement with UK to Explore AI Public Service Applications
• OpenAI Steps Up UK Presence, Collaborates with Government on AI

字节跳动发布70亿参数Seed-X小模型，支持28语种高效翻译

简报：

• 字节跳动Seed团队近日开源了多语言翻译模型Seed-X，拥有70亿参数，支持英语、中文、日语等28种语言的双向翻译。
• Seed-X基于Mistral架构，专注于翻译任务优化，表现优异，能媲美Gemini-2.5、Claude-3.5和GPT-4等顶级大模型。
• 团队采用创新训练策略和数据处理流程，提升了模型多语种翻译能力，并通过MIT协议和Hugging Face平台开源，便于开发者应用。

相关链接：

• https://www.aibase.com/zh/news/19850
• Seed-X项目主页

ByteDance Releases 7 Billion-Parameter Seed-X Small Model, Supporting Efficient Translation in 28 Languages

Briefing:

• ByteDance's Seed team recently open-sourced Seed-X, a multilingual translation model with 7 billion parameters, supporting bidirectional translation across 28 languages, including English, Chinese, and Japanese. 🌍
• Built on the Mistral architecture, Seed-X is optimized for translation tasks, demonstrating outstanding performance comparable to top-tier large models like Gemini-2.5, Claude-3.5, and GPT-4. ✨
• The team employed innovative training strategies and data processing workflows to enhance the model's multilingual translation capabilities. It has been open-sourced under the MIT License and is available on the Hugging Face platform, facilitating developer adoption. 🚀

Related Links:

• https://www.aibase.com/zh/news/19850
• Seed-X Project Page

科大讯飞推出支持本地大模型的智能办公本X5

简报：

• 2025年7月22日，科大讯飞正式发布全球首款搭载本地大模型的墨水屏智能办公设备——讯飞智能办公本X5；
• 该产品首次将自主研发的全国产云端大模型部署于终端本地，内置8核CPU与9T NPU算力；
• 办公本X5可在本地离线实现语音转写、会议纪要生成等核心AI功能，解决网络受限及信息安全等需求。

相关链接：

• 36氪报道

iFLYTEK Launches X5 Smart Office Notebook with On-Device Large Language Model

Brief:

• On July 22, 2025, iFLYTEK officially unveiled the iFLYTEK Smart Office Notebook X5, the world's first e-ink smart office device equipped with a local large language model. 🚀
• For the first time, this product deploys iFLYTEK's self-developed, originally cloud-based large model directly on the device, featuring a built-in 8-core CPU and 9 TOPS NPU computing power. 💪
• The X5 Office Notebook can perform core AI functions such as speech-to-text transcription and meeting minute generation locally and offline, addressing needs related to network limitations and information security. 🔒

Related Links:

• 36Kr Report

谷歌 Gemini 高级版斩获 2025 数学奥林匹克金牌

简报：

• 谷歌 Gemini 高级版以卓越的数学能力斩获 2025 年国际数学奥林匹克（IMO）金牌，成为数学奥赛领域的全新挑战者，展示出 AI 在高端数学竞赛中的突破性表现。

相关链接：

• 相关报道1
• 相关报道2

Google Gemini Advanced Wins 2025 International Mathematical Olympiad Gold Medal 🥇

Briefing:

• Google Gemini Advanced has secured a gold medal at the 2025 International Mathematical Olympiad (IMO) with its exceptional mathematical prowess, emerging as a new challenger in the field of math Olympiads and demonstrating AI's breakthrough performance in high-level mathematical competitions. 🤖✨

Related Links:

• Related Report 1
• Related Report 2

Pika发布AI视频特效APP，自拍一键转化多风格创意大片

简报：

• Pika正式推出面向C端用户的AI视频特效APP，用户可通过上传自拍照生成多样化的视频风格，包括卡通、电影、复古等，降低了创作门槛。
• 新应用支持音频同步表演、场景与风格定制、自拍与视频片段混剪等功能，实现个性化创作与内容混合。
• APP内置智能脚本生成功能，支持通过关键词自动生成视频脚本，提升短视频内容产出效率。
• 此前Pika以B端服务为主，此次进军C端市场，让普通用户也能便捷体验AI视频创作，目前已成为内容创作者热门选择。
• 该APP生成的视频画质高清、适配主流短视频平台，推动AI内容创作方式普及。

相关链接：

• AIbase新闻解读

Pika Launches AI Video Effects App: One-Tap Selfie Transformation into Multi-Style Creative Masterpieces 🚀

Briefing:

• Pika has officially launched its AI video effects app for consumer (C-end) users, allowing them to upload selfies and generate diverse video styles, including cartoon, cinematic, and vintage, thus lowering the creative barrier. 🎉
• The new application supports features such as audio-synced performance, scene and style customization, and mixed editing of selfies and video clips, enabling personalized creation and content blending.
• The app features a built-in intelligent script generation function, supporting automatic video script generation via keywords to enhance the efficiency of short video content production.
• Previously, Pika primarily focused on B-end services; this venture into the C-end market allows ordinary users to easily experience AI video creation and has already become a popular choice among content creators. ✨
• Videos generated by this app are high-definition and compatible with mainstream short video platforms, promoting the widespread adoption of AI-powered content creation methods.

Related Links:

• AIbase News Interpretation