Microsoft Build: company creates “open agent network” for autonomous systems

Microsoft presented a new strategy for working with artificial intelligence, focusing on creating a so-called “open agent network.” At the annual Build conference, the company announced dozens of tools and platforms for developers, designed to create autonomous systems capable of making decisions and performing tasks with minimal human involvement.

Special attention was drawn to the Windows AI Foundry platform. This is a rebranding and expansion of the Windows Copilot Runtime service, launched in May last year. Microsoft positions it as a “unified platform for local artificial intelligence development.” In simple terms, it’s a way to configure, optimize, and deploy the artificial intelligence models that underlie applications for Windows.

Windows AI Foundry is capable of automatically detecting the hardware of a Windows computer. And downloading the necessary software components to run a specific artificial intelligence model. The platform will also keep these components up to date as new devices are released. And provide tools to simplify the model preparation process.

An important part of the initiative was Foundry Local — a new service, I quote, “bringing the power of artificial intelligence models to client devices.” Foundry Local, which supports both Windows and macOS, includes a set of modules for running artificial intelligence models directly on the device, using ONXX Runtime. This is an open-source project aimed at accelerating artificial intelligence on various platforms.

Well, Microsoft seems to want to cover all levels of the artificial intelligence stack. From optimizing model operation at the operating system level to creating development tools. The strategy is the same as with the Windows operating system. This is forming the dependence of developers on the Microsoft ecosystem.