简报: OpenAI正式发布了其最新开源项目Codex CLI,这是一款轻量级的代码智能体工具。该工具的发布迅速引起了极大的关注,仅在短短5小时内,Codex CLI便突破了5000颗星的评价,预计今天有望突破1万颗星,迅速成为Github上的一颗新星。Codex CLI具备多种强大的功能,能够根据用户输入的提示自动生成代码文件、运行代码、安装缺失的依赖并实时展示结果。开源地址:https://github.com/openai/codex?tabreadme-ov-file
AI 圈又热闹了!OpenAI 近日发布了新一代推理模型 o3 和 o4-mini,最抓人眼球的莫过于它们学会了"看图说话"。这可不是简单的图像识别,而是像人类工程师一样,能对设计图纸进行推敲、理解。OpenAI 的 CEO Sam Altman 对此赞不绝口,直呼它们"接近天才水平"。
The AI world is buzzing again! OpenAI recently released its new generation of reasoning models, o3 and o4-mini. What's most eye-catching is that they've learned to "describe images." 🤩 This isn't just simple image recognition; it's like human engineers being able to scrutinize and understand design blueprints. 🤯 OpenAI CEO Sam Altman raved about them, calling them "close to genius level." 🧠✨
简报:
人工智能再次刷新科学探索的边界!一位名为 Weiguo Yin 的华人学者,利用 OpenAI 的 o3-mini-high 模型,成功"啃"下了一个困扰学界长达半个世纪的数学硬骨头,完成了首个由 AI 辅助的非平凡数学定理证明。这事儿,有点意思,或许真能打开物理学研究的新局面。
谷歌最近扔出个重磅炸弹——DolphinGemma,一个能破译海豚叫声的AI模型。等等,这难道不是科幻电影里的情节吗?
Google recently dropped a bombshell – DolphinGemma, an AI model that can decipher dolphin sounds. Wait, isn't this something straight out of a sci-fi movie? 🤯