- Current AI agents can connect but cannot “think together,” creating a major barrier for next-generation AI systems.
- According to Vijoy Pandey from Cisco, agents lack shared context and synchronization of “meaning” when working together.
- Each agent operates almost from scratch in every run, without sharing memory or accumulated insights.
- The concept of “shared cognition” is the next step: allowing AI to collaborate on solving new problems without human intervention.
- This is similar to human evolution: from individuals → communication → collective intelligence.
- Cisco is developing new protocols like SSTP, LSTP and CSTP to transmit “cognitive states” between agents.
- These protocols allow sharing semantics, latent space, and compressed data to increase coordination efficiency.
- The goal is to build an “internet of cognition” – infrastructure for large-scale distributed intelligence.
- In reality, Cisco has deployed over 20 agents to automate processes, reducing processing time from hours to seconds.
- Agents also help reduce up to 80% of errors in Kubernetes systems, showing the great potential of effective coordination.
📌 AI is moving towards a new stage: not just being individually smart but needing to “collaborate like humans.” However, the lack of shared context and intent prevents agents from creating true collective intelligence. Efforts like the “internet of cognition” could unlock the next leap, where AI not only works faster but also thinks together to solve complex problems.
