When earthquakes hit, they often come with little warning. Each year, about 500,000 earthquakes ripple through the Earth—some are felt, many aren’t, but all are part of a complex and dynamic system that geologists are trying to understand. Earthquakes are…
Empowering patients and healthcare consumers in the age of generative AI
| Peter Lee, Dave deBronkart, and Christina Farr
Evangelist for patient empowerment Dave deBronkart and Manatt Heath’s Christina Farr discuss how generative AI is redefining healthcare by empowering patients, challenging traditional care models, and creating new opportunities for innovation and collaboration.
硅基世界的“缘分”——系统与人工智能携手进化
作者：周礼栋 编者按：在 NeurIPS 2024 大会上，微软亚洲研究院院长周礼栋博士发表了题为“硅基世界的‘缘分’——系统与人工智能携手进化”的主题演讲。他提出了一个前瞻性的观点：未来人工智能的突破，不仅依赖于模型和算法本身，更有赖于与计算机系统的深度融合。围绕“效率（Efficiency）、可信（Trust）、融合（Infusion）”三大关键词，周礼栋博士阐述了系统与人工智能之间如何实现双...
ICLR上新 | 探索大模型优化、信息检索与记忆构建的创新之路
编者按：欢迎阅读“科研上新”栏目！“科研上新”汇聚了微软亚洲研究院最新的创新成果与科研动态。在这里，你可以快速浏览研究院的亮点资讯，保持对前沿领域的敏锐嗅觉，同时也能找到先进实用的开源工具。 4月24日至4月28日，全球最负盛名的深度学习盛会之一 ICLR 大会将在新加坡举办。我们将通过两期“科研上新”为大家带来多篇微软亚洲研究院入选 ICLR 2025 的精选论文解读。第一期的研究工作将涵盖大模...
In the news | VentureBeat
When AI reasoning goes wrong: Microsoft Research shows more tokens can mean more problems
Large language models (LLMs) are increasingly capable of complex reasoning through “inference-time scaling,” a set of techniques that allocate more computational resources during inference to generate answers. However, a new study from Microsoft Research reveals that the effectiveness of these…
Engagement, user expertise, and satisfaction: Key insights from the Semantic Telemetry Project
| Amber Hoak, Scott Counts, Kate Lytvynets, David Tittsworth, Siddharth Suri, Nirupama Chandrasekaran, Ben Cutler, and Weiwei Yang
Semantic Telemetry Project data show that people who use AI for more professional and complex tasks are more likely to keep using the tool and to use it more often. Novice AI users engage in simpler tasks, but their usage…
StreamMind：解锁全帧率流媒体视频实时理解新路径
当今基础大模型的能力提升催生了诸多真实世界应用，如智能助手等。这类助手通过智能眼镜等智能设备时刻与用户相伴，持续感知和理解用户所在环境、发生的事件以及用户指令，并据此实时主动地给用户提供帮助。比如导盲助手，在用户要过马路时会输出“稍等现在有车经过”，“现在可以安全通过”等帮助提示。除了依赖大语言模型的世界知识和推理能力，此类智能助手更依赖对持续输入视频流的理解能力，以及可以在相关事件发生时立即发出...
Debug-gym: an environment for AI coding tools to learn how to debug code like programmers
| Eric Yuan, Morgane Moss, Charbel Feghali, Chinmay Singh, Darya Moldavskaya, Drew MacPhee, Lucas Caccia, Matheus Pereira, Minseon Kim, Alessandro Sordoni, and Marc-Alexandre Côté
Developers spend a lot of time debugging code. Learn how debug-gym can equip AI agents to help, enabling them to set breakpoints, navigate the codebase, and print runtime variable values on demand, so they better understand the code and its…
In the news | LinkedIn Article
Unlocking AI’s Global Potential: Progress, Productivity, and Workforce Development
Stanford HAI Index highlights transformative developments in artificial intelligence that carry profound implications for societies worldwide—especially in regions across the Global South [1]. As we explore these insights, we recognize that AI is transforming industries, creating new opportunities, and driving economic…