Microsoft aims to create an 'agentic web' where AI agents from different companies work together using the Model Context Protocol (MCP).
MCP, an open-source standard created by Anthropic, allows AI agents to collaborate to improve their accuracy through reinforcement learning.
Agentic AI, specialized in specific tasks like cybersecurity and customer support, heavily relies on training data and reinforcement learning for optimal performance.
The potential for AI agents to autonomously interact within an 'agentic web' raises concerns regarding negative feedback loops and the need for preventive mechanisms.