By Stephen Nellis
REDMOND, Washington (Wiseova) – Microsoft anticipates a future where AI representatives from various companies will be able to collaborate seamlessly with one another and retain enhanced recollections of past communications, according to the firm’s top technologist. This statement was made prior to the commencement of the corporation’s yearly gathering for software developers.
Microsoft is hosting its Build conference in Seattle on May 19, and industry experts anticipate the unveiling of new development tools focused on artificial intelligence at this event.
At Microsoft’s headquarters in Redmond, Washington, prior to the conference, Chief Technology Officer Kevin Scott addressed reporters and analysts stating that the company aims to promote the implementation of industry-wide standards. These standards would enable agents created by various manufacturers to work together seamlessly. According to Scott, these agents refer to artificial intelligence systems capable of performing particular duties independently, like resolving a software issue.
Scott mentioned that Microsoft is supporting a technology known as Model Context Protocol (MCP), which is an open-source protocol initially presented by Google-supported Anthropic. According to Scott, MCP could lead to the development of an "agentic web," much like how hypertext protocols facilitated the expansion of the internet during the 1990s.
Scott stated, "This implies that your creativity will steer how the agentic web evolves, rather than having control solely reside with a few corporations that merely happened upon some of these issues initially."
Scott additionally mentioned that Microsoft aims to assist AI agents in retaining better recollections of tasks requested by users. He pointed out that currently, much of their work seems quite transactional.
However, enhancing an AI agent’s memory is very expensive due to the increased computational requirements. Microsoft is concentrating on a novel method known as structured retrieval augmentation. This technique involves the agent capturing brief segments from every exchange in a conversation with a user, thereby forming a map of sorts for reference purposes.
As Scott explained, this is fundamental to training a biological brain—you don’t have to start from scratch each time you face a specific challenge.
(Reported by Stephen Nellis; Edited by Paul Simao)
0 Comments