Cisco Executive: AI Agents Lack Shared Cognition, Need New Protocols
Vijoy Pandey, SVP and GM of Outshift by Cisco, identifies a critical bottleneck in artificial intelligence: while AI agents can connect, they cannot currently think together due to a lack of semantic alignment and shared context. To address this, Pandey proposes the development of an 'internet of cognition' aimed at achieving distributed super intelligence. His team is creating new infrastructure layers, including specific protocols like Semantic State Transfer Protocol (SSTP), Latent Space Transfer Protocol (LSTP), and Compressed State Transfer Protocol (CSTP), alongside cognition engines to synchronize states across endpoints. This approach mirrors human cognitive evolution, moving from individual intelligence to collective coordination. Beyond theoretical frameworks, Cisco has implemented existing AI agent technologies within its site reliability engineering teams. By automating workflows such as CI/CD pipelines and Kubernetes deployments, over 20 agents have reduced deployment times from hours to seconds and resolved 80% of related issues. Pandey emphasizes that while these advancements are significant, AI must be integrated with deterministic code to solve problems effectively, rather than being applied indiscriminately.
Wire timeline
Cisco Executive: AI Agents Lack Shared Cognition, Need New Protocols
Vijoy Pandey, SVP and GM of Outshift by Cisco, identifies a critical bottleneck in artificial intelligence: while AI agents can connect, they cannot currently think together due to a lack of semantic alignment and shared context. To address this, Pandey proposes the development of an 'internet of cognition' aimed at achieving distributed super intelligence. His team is creating new infrastructure layers, including specific protocols like Semantic State Transfer Protocol (SSTP), Latent Space Transfer Protocol (LSTP), and Compressed State Transfer Protocol (CSTP), alongside cognition engines to synchronize states across endpoints. This approach mirrors human cognitive evolution, moving from individual intelligence to collective coordination. Beyond theoretical frameworks, Cisco has implemented existing AI agent technologies within its site reliability engineering teams. By automating workflows such as CI/CD pipelines and Kubernetes deployments, over 20 agents have reduced deployment times from hours to seconds and resolved 80% of related issues. Pandey emphasizes that while these advancements are significant, AI must be integrated with deterministic code to solve problems effectively, rather than being applied indiscriminately.
VentureBeat