In the run-up to Microsoft's annual developer conference, Kevin Scott, Microsoft's Chief Technology Officer, revealed an exciting vision: in the future, artificial intelligence (AI) agents from different companies will be able to collaborate seamlessly and remember their interactions with each other more effectively. Scott stated at Microsoft's headquarters in Redmond, Washington, that the company is committed to promoting the establishment of industry standards to enable AI agents from various manufacturers to work together.
Microsoft will support a standardized communication protocol called Model Context Protocol (MCP). This protocol allows AI applications to communicate smoothly with external data sources and tools, giving AI models a universal way to access and use information beyond their own training data, thereby enhancing the capabilities and efficiency of AI systems. Scott compared MCP to the role of hypertext in the development of the internet in the 1990s, implying that this technology could form a collaborative network of AI agents.
Furthermore, Scott mentioned that Microsoft is exploring solutions to help AI agents better remember user requests. He pointed out that existing interactions often appear very isolated and lack coherence. However, improving the memory capabilities of AI systems is costly, as it requires additional computational resources. To address this, Microsoft is focusing on a new approach called structured retrieval augmentation. This method enables AI agents to extract short segments from user interactions, creating a summary of the discussion to help them better understand users' needs in future interactions.
As AI technology continues to evolve, Microsoft's efforts will not only promote collaboration within the tech industry but also provide users with smarter AI experiences. It is believed that future AI agents will no longer be isolated individuals but an interconnected and mutually understanding intelligent network.