The path to artificial superintelligence
MIT Technology Review AI MIT Technology Review Insights
Cisco's Outshift proposes a multi-agent AI system architecture called the 'Internet of Cognition' that enables independent AI agents across different domains to coordinate and reason together through semantic, connectivity, and security layers. Current multi-agent systems have failure rates between 41% and 87%, but Outshift's approach using open protocols like AGNTCY and coordination mechanisms like Mycelium raised decision-making success rates to 93% in internal testing. Organizations can begin experimenting with cross-functional multi-agent workflows to enable intelligence to scale horizontally across systems rather than just vertically through larger models.
Why it matters
Imagine a healthcare system made up of multiple AI agents: one that manages symptom assessment, another scheduling, a third insurance, and a fourth pharmacy. Each is an expert in its domain. But they all have their own distinct knowledge and objectives. Today they can exchange data, but they are not yet able to actually coordinate…
