Redmond: Microsoft imagines a future in which artificial intelligence agents of any company can work together with agents from other companies and have better memories of their interactions, said their main technologist on Sunday before the company’s annual software developers conference.
Microsoft celebrates its compilation conference in Seattle on May 19, where analysts expect the company to present their latest tools for developers who build AI systems.
Speaking at the Microsoft headquarters in Redmond, Washington, before the conference, said technology director Kevin Scott to journalists and analysts that the company focuses on helping to stimulate the adoption of standards throughout the technology industry that will allow agents of different manufacturers to collaborate. Agents are AI systems that can perform specific tasks, such as fixing a software error, on their own.
Scott said Microsoft is supporting a technology called Model Context Protocol (MCP), an open source protocol introduced by anthropic backed by Google. Scott said MCP has the potential to create an “agent network” similar to the way in which hypertext protocols that helped spread Internet in the 1990s.
“It means that your imagination drives what the agent network becomes, not just a handful of companies that see some of these problems first,” Scott said.
Scott also said that Microsoft is trying to help AI agents have better memories of the things that users have asked them to do, noting that, so far, “most of what we are building feels very transactional.”
But making the memory of an agent for a lot of money because it requires more computer power. Microsoft is focusing on a new approach called structured recovery increase, where an agent extracts short bits from each turn in a conversation with a user, creating a road map to what was discussed.
“This is a central part of how a biological brain trains: I don’t force everything in your head every time you need to solve a particular problem,” Scott said.
