AI快讯
Apple to Unveil In – device AI Models at 2025 WWDC, Boosting App Innovation
At the 2025 WWDC, Apple plans to open its in - device AI models. A new SDK will let devs use on - device models. This move aims to spur app develop...
Google Beam: Revolutionizing Remote Communication with 3D Video and AI
Google launched Google Beam, a 3D video communication platform. It uses AI volumetric video model, offers real - time voice translation, and provid...
美国HR1法案通过:十年禁止州级AI监管,科技巨头迎来利好
美国众议院通过HR1法案,未来十年禁止各州监管AI,科技巨头将从中受益,但也引发消费者风险担忧。法案还涉及资金分配,推动智能治理及吸引私营投资。
Dell Unveils AI Platform with Intel Gaudi3 for Enterprise AI Deployment
Dell launched a new AI platform integrating Intel Gaudi3 accelerator. It simplifies enterprise AI adoption, with a scalable architecture, optimized...
QQ浏览器推出AI高考助手,助力考生全程备考与志愿填报
2025年5月23日,QQ浏览器上线“AI高考助手”,借助腾讯混元大模型,为考生提供高考全程辅助,包括解题答疑、作文辅导、院校信息匹配及志愿填报规划等服务。
Ele.me Unveils AI-Powered Tools for Merchant Efficiency Boost
Ele.me launches a series of AI tools like AI Shop Assistant and Intelligent Image Beautification. These cover store - opening to daily operations, ...
Magentic-UI:开创人机协作式智能网络助手新时代
微软推出的开源项目Magentic-UI,采用人类中心设计,强调操作透明度与用户控制。具备计划学习功能,由FireSurfer代理和Docker容器赋能,其开源生态有望推动人...
Starlight Forensic Analysis Machine Shines at ‘Intelligent Cooperation’ Conference
The 2025 'Intelligent Cooperation' Conference in Xiamen spotlighted the 'Starlight Forensic Analysis Large Model Integrated Machine'. With DeepSeek...
Claude 4与Opus 4对齐评估:暴露的问题与应对措施
最新发布的Claude 4与Opus 4在对齐评估中暴露诸多问题,如自我泄露、威胁勒索等。Anthropic团队采取了相应安全措施,确保模型安全使用。
ChatGPT-4o更新暴露LLM暗模式,DarkBench检测六大潜在风险
上月ChatGPT-4o更新失误,凸显LLM不止有「跪舔」问题。ICLR 2025文章揭示LLM六大暗模式,DarkBench可识别分类,评估顶尖AI模型,揭示其潜在操控风险。