Research Focus: Week of April 7, 2025
In this issue: We introduce a new dataset designed to assist renewable energy infrastructure planners, a new method for denoising MRI imagery, and an AI tool for analyzing distant galaxies. Check out our latest research and other updates.
从能力到效率，多管齐下提升大模型智能体系统的智能“密度”
作者：杨玉庆 大模型的发展已成为推动软件形态巨变的关键因素之一，它正在催生一种全新的软件形态，即目前我们所理解的智能体（AI agent）或智能体系统（agentic system）。 不难想象，智能体系统将会是大模型应用的普遍形式。它的性能表现和使用成本，对于大模型所能惠及的范围具有决定性作用。因此，我们尝试以这样一种思路来优化智能体系统：将其视为一个“数据加工”流程，并尝试在不同环节为工作量做...
MaaG：迈向世界模型，解决生成式游戏中的数值与时空一致性难题
世界模型（world models）是人工智能领域的一个重要概念，旨在通过模拟虚拟世界中主体行为的演变，实现高度逼真的互动体验。这种模型不仅可以为游戏和互动媒体的生成带来革命性的变化，也将为人工智能在复杂环境中的应用提供新的可能性。其中，生成式游戏（generative games）作为构建世界模型的关键途径，备受关注。例如，微软提出的 MUSE 能够用神经网络生成游戏《嗜血边缘（Bleeding...
In the news | NEJM Catalyst
Jim Weinstein discusses health care challenges in U.S. rural areas with NEJM Catalyst
Disparities include limited access to care, rising mortality rates, and severe staffing shortages.
Awards | Ministry of Education, Culture, Sports, Science and Technology - Japan
Yasuyuki Matsushita receives MEXT Minister of Education, Culture, Sports, Science and Technology Award
Recipients have achieved remarkable results in research and development and promotion of understanding of science and technology.
A key challenge, and opportunity, of large language models (LLMs) is bridging the gap between their training data and the vast amount of unfamiliar information they encounter in real-world applications. Successfully navigating this divide could unlock a new era of…
Today we are making available an interactive real-time gameplay experience in Copilot Labs. Head over to this link (opens in new tab) to play an AI rendition of Quake II gameplay, powered by Muse. 5 second video of Quake II…
Real-world healthcare AI development and deployment—at scale
| Peter Lee, Matthew Lungren, and Seth Hain
Microsoft’s Dr. Matthew Lungren and Epic’s Seth Hain discuss the challenges and opportunities of leveraging generative AI for enhanced patient care and improved clinical documentation and recordkeeping at scale—plus what’s next for the technology in the field.
VidTok introduces compact, efficient tokenization to enhance AI video processing
| Tianyu He, Junliang Guo, and Jiang Bian
The VidTok method can enable AI systems to process and generate videos more effectively. Compact tokenization reduces computational costs while maintaining video quality across a potentially diverse set of applications.