AI科技热点早报 2025-05-19 8:00

✅新鲜，及时的AI科技热点新闻，覆盖全网主流科技媒体源。

✅每日上午8点定时发布，敬请关注。

01 OpenAI发布编程Agent Codex，支持多任务处理

• 内容源：OpenAI 发布编程 Agent 产品 Codex 🏷️ifanr
• 摘要：OpenAI 正式发布了一款编程 Agent 产品「Codex」。OpenAI 介绍，Codex 是一款支持并行处理多个任务的云端编程 Agent，能够提供如编程功能、回答代码库的问题、修复错误等功能。据悉，Codex 基于 codex-1 模型驱动，OpenAI 方面表示这一模型由 o3 模型针对编程进行优化而得来。

• 内容源：刚刚，OpenAI丢出最强编程智能体Codex！倒反天罡——新上线功能竟是微软Copilotb鼻祖？ 🏷️51cto
• 摘要：OpenAI 一边刚刚收购了浏览器创业公司 Windsurf，一边又在编程领域持续加码 Codex 智能体，不难看出，他们正在为“AI 全栈开发环境”铺路。

• 内容源：LLMs Struggle with Real Conversations: Microsoft and Salesforce Researchers Reveal a 39% Performance Drop in Multi-Turn Underspecified Tasks 🏷️MarkTechPost
• 摘要：Conversational artificial intelligence is centered on enabling large language models (LLMs) to engage in dynamic interactions where user needs are revealed progressively. These systems are widely deployed in tools that assist with coding, writing, and research by interpreting and responding to natural language instructions. The aspiration is for these models to flexibly adjust to changing […]The postLLMs Struggle with Real Conversations: Microsoft and Salesforce Researchers Reveal a 39% Performance Drop in Multi-Turn Underspecified Tasksappeared first onMarkTechPost.

02 Meta推出LlamaFirewall，增强AI安全防护

• 内容源：Meta 推出开源框架 LlamaFirewall，强化 AI Agent安全防护 🏷️InfoQ
• 摘要：无

03 AWS开源Strands Agents SDK，简化AI代理开发

• 内容源：AWS Open-Sources Strands Agents SDK to Simplify AI Agent Development 🏷️MarkTechPost
• 摘要：Amazon Web Services (AWS) has open-sourced its Strands Agents SDK, aiming to make the development of AI agents more accessible and adaptable across various domains. By following a model-driven approach, the Strands Agents SDK abstracts much of the complexity behind building, orchestrating, and deploying intelligent agents—making it easier for developers to build tools that plan, […]The postAWS Open-Sources Strands Agents SDK to Simplify AI Agent Developmentappeared first onMarkTechPost.

04 谷歌推出LightLab，提升图像光照控制能力

• 内容源：Google Researchers Introduce LightLab: A Diffusion-Based AI Method for Physically Plausible, Fine-Grained Light Control in Single Images 🏷️MarkTechPost
• 摘要：Manipulating lighting conditions in images post-capture is challenging. Traditional approaches rely on 3D graphics methods that reconstruct scene geometry and properties from multiple captures before simulating new lighting using physical illumination models. Though these techniques provide explicit control over light sources, recovering accurate 3D models from single images remains a problem that frequently results in […]The postGoogle Researchers Introduce LightLab: A Diffusion-Based AI Method for Physically Plausible, Fine-Grained Light Control in Single Imagesappeared first onMarkTechPost.

05 英伟达计划在上海设立研发中心，提升AI芯片竞争力

• 内容源：英伟达将在上海设立研发中心 🏷️ifanr
• 摘要：消息称英伟达计划在上海设立研究中心据《金融时报》消息，英伟达计划在上海建立一座研发中心，旨在让其在中国保持 AI 芯片领域的竞争力报道指出，由于此前美国出口管控的收紧，英伟达在华销售额已出现下滑。

• 内容源：Nvidia CEO Sees No Evidence of AI Chip Diversion Into China 🏷️Bloomberg
• 摘要：Nvidia Corp. Chief Executive Officer Jensen Huang expressed confidence in his company’s trade partners and said there’s no evidence of its prized semiconductors being diverted to the Chinese market.

06 谷歌推出DolphinGemma，探索海豚语言的AI工具

• 内容源：谷歌推出 DolphinGemma：探索海豚语言的 AI 工具 🏷️InfoQ
• 摘要：无

08 DeepSeek-AI研究高效语言建模，降低硬件开销

• 内容源：This AI paper from DeepSeek-AI Explores How DeepSeek-V3 Delivers High-Performance Language Modeling by Minimizing Hardware Overhead and Maximizing Computational Efficiency 🏷️MarkTechPost
• 摘要：The growth in developing and deploying large language models (LLMs) is closely tied to architectural innovations, large-scale datasets, and hardware improvements. Models like DeepSeek-V3, GPT-4o, Claude 3.5 Sonnet, and LLaMA-3 have demonstrated how scaling enhances reasoning and dialogue capabilities. However, as their performance increases, so do computing, memory, and communication bandwidth demands, placing substantial strain […]The postThis AI paper from DeepSeek-AI Explores How DeepSeek-V3 Delivers High-Performance Language Modeling by Minimizing Hardware Overhead and Maximizing Computational Efficiencyappeared first onMarkTechPost.

09 苹果与阿里巴巴合作引发美国官员担忧

• 内容源：ChatGPT 或将支持 MCP；字节跳动调整福利政策；洛杉矶奥运会将引入「空中出租车」 🏷️极客公园
• 摘要：OpenAI 计划在阿布扎比建设的数据中心将比摩纳哥更大据彭博社报道，OpenAI 准备帮助在阿布扎比开发一个容量高达 5 千兆瓦的数据中心园区，该公司将成为该项目的主要租户，该园区可能成为全球最大的人工智能基础设施项目之一。据报道，该设施占地面积高达 10 平方英里，耗电量相当于五座核反应堆，远超 OpenAI 及其竞争对手宣布的任何现有 AI 基础设施。OpenAI 在美国的第一座星际之门园区（已在德克萨斯州阿比林开发）预计将达到 1.2 千兆瓦，而这家位于中东的园区的容量将是该园区的四倍以上。（来源：cnBeta）

10 ICML 2025提出基于RAG的高分辨率图像感知框架

• 内容源：ICML 2025 Spotlight｜南洋理工陶大程教授团队等提出基于RAG的高分辨率图像感知框架，准确率提高20% 🏷️机器之心
• 摘要：无

11 谷歌AlphaEvolve突破算法效率限制

• 内容源：打破56年数学铁律！谷歌AlphaEvolve自我进化实现算法效率狂飙 🏷️量子位
• 摘要：相当于AlphaGo打败人类的“神之一手”

12 微软和Salesforce研究显示LLM在对话中表现不佳

• 内容源：LLMs Struggle with Real Conversations: Microsoft and Salesforce Researchers Reveal a 39% Performance Drop in Multi-Turn Underspecified Tasks 🏷️MarkTechPost
• 摘要：Conversational artificial intelligence is centered on enabling large language models (LLMs) to engage in dynamic interactions where user needs are revealed progressively. These systems are widely deployed in tools that assist with coding, writing, and research by interpreting and responding to natural language instructions. The aspiration is for these models to flexibly adjust to changing […]The postLLMs Struggle with Real Conversations: Microsoft and Salesforce Researchers Reveal a 39% Performance Drop in Multi-Turn Underspecified Tasksappeared first onMarkTechPost.

13 奇瑞计划在英国建汽车厂，扩展市场布局

• 内容源：SWE-Bench Performance Reaches 50.8% Without Tool Use: A Case for Monolithic State-in-Context Agents 🏷️MarkTechPost
• 摘要：Recent advancements in LM agents have shown promising potential for automating intricate real-world tasks. These agents typically operate by proposing and executing actions through APIs, supporting applications such as software engineering, robotics, and scientific experimentation. As these tasks become more complex, LM agent frameworks have evolved to include multiple agents, multi-step retrieval, and tailored scaffolding […]The postSWE-Bench Performance Reaches 50.8% Without Tool Use: A Case for Monolithic State-in-Context Agentsappeared first onMarkTechPost.

感谢您的阅读。原创不易，如您觉得有价值，请点赞，关注。