Izzo: You're listening to Exploring Next, episode 294. Today, we're talking about the next bottleneck in AI: whether agents can think together.
Boone: This matters right now because we're seeing more and more AI agents being deployed, but they're not really working together seamlessly.
Izzo: Exactly. And it's not just about connecting them, it's about creating shared cognition. Vijoy Pandey from Cisco describes it as the 'internet of cognition'.
Boone: That's a great point. So, what is shared cognition, and how do we achieve it?
Izzo: Well, shared cognition refers to AI agents working together to solve new problems without human intervention. It's like human intelligence, where we work together to achieve a common goal.
Boone: And it requires codifying intent, context, and collective innovation as rules, APIs, and capabilities. It's a horizontal distributed assistance problem.
Izzo: Right. And Cisco is working on developing new protocols to support this, like SSTP, LSTP, and CSTP.
Boone: SSTP operates at the language level, analyzing semantic communication. LSTP can transfer the entire latent space of one agent to another. And CSTP handles compression, grounding only targeted variants.
Izzo: That's really interesting. And what about the practical applications of this? How is Cisco using AI agents to solve real-world problems?
Boone: Well, they've achieved tangible results with existing AI capabilities, like automating end-to-end workflows and improving efficiency for their site reliability engineering team.
Izzo: Okay, so what can our listeners go research or try installing to get hands-on with this?
Boone: I'd recommend checking out the Ripple Effect Protocol, and looking into the SSTP, LSTP, and CSTP protocols. Maybe even try implementing some of these concepts in a personal project or a weekend hackathon.
Izzo: Alright, that sounds like a great place to start. Thanks for tuning in to this episode of Exploring Next, and we'll catch you on the next one.