IT之家 October 8 news: At a launch event held in San Francisco, USA, at 10 a.m. Pacific Time on October 7 (1 a.m. Beijing Time on October 8), Microsoft proposed the concept of "hybrid intelligence," emphasizing a hybrid architecture that combines cloud AI with local computing power,on Windows 11 to actually process user tasks.
According to a blog post cited by IT之家, Windows AI is no longer limited to a single cloud call or on-device execution; it will coordinate between local and cloud depending on task type, device performance, and security requirements: privacy-sensitive tasks that require fast response can rely more on local processing, while complex models and high computing demands are handed off to the cloud.
Windows may become a key entry point for this model: after the user states a goal, Copilot and AI agents are responsible for understanding context, invoking relevant tools, and advancing the task, while the user retains control over key steps, permissions, and final results.
This concept echoes Microsoft's previously proposed "frontier ecosystem." Microsoft CEO Satya Nadella views trust, intelligence, and "AI woven into human ambition" as three pillars, and calls Agent 365, Microsoft IQ, and Copilot Microsoft's "key pivot points."
Under this framework, hybrid intelligence emphasizes redesigning the way people collaborate with Windows, applications, and agents. Microsoft also highlighted the importance of hybrid architecture and local—cloud coordination: on-device computing can handle tasks requiring low latency and privacy protection, while the cloud provides stronger models and computing power. Combined with controlled execution environments such as Execution Containers, AI agents are expected to perform more complex operations with the necessary permissions.
At this event, Microsoft Senior Technical Program Manager Kayla Cinnamon demonstrated GitHub Copilot, which automatically selects local models rather than blanketly calling cloud models for processing.