编者按:如今,大模型和生成式人工智能的爆发正把计算推向前所未有的复杂境地:硬件以惊人的速度迭代,应用负载瞬息万变,而传统系统却仍被束缚在过往的固化规则中。面对这种“上下皆动”的挑战,刘璟和同事正试图在人工智能的不可预测性与系统的刚性之间,为操作系统注入全新的自主适应能力。 在计算机科学的世界里,系统一直扮演着沉默的基石角色,它托举着上层应用的高效运行,却深埋于底层硬件的复杂逻辑之中。对微软亚洲研究...
On Second Thought is a video series built around the questions everyone’s asking about AI. With expert voices from across Microsoft, we break down the tension and promise of this rapidly changing technology, exploring what’s evolving and what’s possible.
In the news | World Economic Forum
Food systems, on land and at sea, are foundational to global economies and human flourishing. They are sources of livelihood for millions of people and provide nutrition for everyone on the planet. Yet today, these systems are under pressure, making…
In the news | Fast Company
From look-alike SUVs to AI interfaces, designers from Figma, Microsoft, Pentagram, and more reveal the objects, places, and experiences they’re working to improve this year.
编者按:数据稀缺是目前限制通用机器人操作大模型进展的重要因素之一。基于遥操的机器人轨迹获取成本高、多样性不足、本体多变,难以获得支撑大模型预训练的海量数据。在人形机器人的发展路径中,人类可被视为最重要的参考形态之一。因此,直接使用真实的人类数据进行预训练,比用机器人数据具有更大的潜力。因为人类的形态和动作模式更稳定、通用,人类数据规模大、多样化且易于获取,更可能达到规模化,以提升模型的泛化能力。...
In the news | Microsoft Corporate Responsibility
Global adoption of artificial intelligence continued to rise in the second half of 2025, increasing by 1.2 percentage points compared to the first half of the year, with roughly one in six people worldwide now using generative AI tools, remarkable progress for a…
In the news | Reuters
January 7 - In the past decade, cloud-scale analytics tools have transformed the digital fight against deforestation. Instead of manual reviews of satellite images taking multiple months, land-use change can today be monitored in near-real time. Some artificial intelligence-enabled tools…
In the news | Tools and Weapons with Brad Smith
Welcome to a special edition of Tools and Weapons, as Brad Smith and Juan M. Lavista Ferres review the results from Microsoft’s newest AI Diffusion Report. These show the state of global AI usage as we look ahead in 2026,…
《AI Next》是微软亚洲研究院推出的一档利用 AI 技术制作的播客,内容聚焦 AI 前沿技术、科研趋势与社会影响。第一季主要围绕当今智能发展的核心议题,探索前沿趋势。 《AI Next》第三期节目,我们将目光投向近年来发展迅速的 Agent(智能体)和 Agentic AI,与微软亚洲研究院首席研发工程师杨玉庆与高级研究员王韵,共同探讨 AI 如何从对话式“助手”演进为能够被人放心委托任务的智...