Key Takeaways
- Microsoft envisions a future where AI assistants from different companies can team up and work together.
- The company also wants these AIs to have much better memories of past conversations and user requests.
- To enable this collaboration, Microsoft is backing a technology called Model Context Protocol (MCP).
- MCP aims to create an “agentic web,” potentially transforming how AIs interact, similar to how the early internet evolved.
- For smarter AI memory, Microsoft is developing a cost-effective method called “structured retrieval augmentation.”
Microsoft is looking towards a future where artificial intelligence agents, regardless of who developed them, can collaborate effectively. The company’s chief technologist shared this vision ahead of its annual software developer conference.
The tech giant is holding its Build conference in Seattle, where analysts anticipate the unveiling of new tools for those building AI systems.
Speaking from Microsoft’s headquarters, Chief Technology Officer Kevin Scott told reporters and analysts the company is focused on encouraging industry-wide standards. These standards would allow AI agents from different creators to work in concert. Agents, in this context, are AI systems designed to perform specific tasks autonomously, such as fixing a software bug.
Microsoft is throwing its support behind a technology called Model Context Protocol (MCP), an open-source protocol initially introduced by Google-backed Anthropic. Scott believes MCP could pave the way for an “agentic web,” drawing parallels to how hypertext protocols fueled the internet’s growth in the 1990s.
“It means that your imagination gets to drive what the agentic web becomes, not just a handful of companies that happen to see some of these problems first,” Scott said, according to Reuters.
Beyond teamwork, Microsoft is also keen on improving how well AI agents remember previous interactions. Scott noted that currently, “most of what we’re building feels very transactional,” meaning AIs often forget past context.
Enhancing an AI’s memory can be expensive because it requires more computing power. To address this, Microsoft is focusing on a novel approach called “structured retrieval augmentation.” This technique involves an agent extracting key pieces of information from each part of a user conversation, effectively creating a summary map of discussed topics.
“This is a core part of how you train a biological brain – you don’t brute force everything in your head every time you need to solve a particular problem,” Scott explained, highlighting a more efficient path for AI memory.
