May 2023
tl;dr: Use LLM to connect various AI models to solve AI tasks.
API-based approach. Using AI to manage AI models, or an AI-based expert systems. The LLM such as ChatGPT acts like a switch.
- Four steps
- Task planning (with LLM)
- Model selection (with LLM)
- Task execution, and return results to LLM
- Response generation
- Summary of technical details
- Questions and notes on how to improve/revise the current work