Tag: AIAgents

Reading Notes | AutoGen – Enabling Next-Gen LLM Applications via Multi-Agent Conversation

Posted on by David Yang

Overview

Qingyun Wu’s Talk

  • The interaction between the users and the AutoGen system is critical for the system to be useful; fully autonomous systems are not trustworthy. For example, if the system does not deliver the outcome the user want, we could not know which step goes wrong.
  • In order for the system to be useful, the base models are important. We cannot use a very weak LM as one of the AI agents.
  • There are currently (as of 2023-11-01) no safety measures to make sure the system do not generate undesirable content.