Building Agents: Multi-Model Strategies and Secure Enclaves
Nothing is loaded from YouTube until you press play, and then without tracking cookies.
About this video
Creating robust agents using multiple models for diverse tasks. Our approach uses secure enclaves on the CPU for confidentiality and verifiability. We discuss NEAR Shade Agents and agent protocols, achieving autonomous agent capabilities. #Agents #MultiModel #SecureEnclaves #AI #MachineLearning #AutonomousAgents #AgentProtocols #DeepLearning #ComputerScience #Tech #NEARprotocol
Transcript
93 wordsCaptions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.
- 0:00
agent is effectively usage of some model maybe multiple models ideally you want multiple models for different use cases and different special things and then some code that glues it that uses tools that calls information that maintains context etc right so this is again we use the same secure enclave now on CPU side to run the agent code that can run models so you can have full confidentiality of the agent itself and verifiability
- 0:25
agent as well can have a token that represents ownership and you know upgradability and whatever else.