> Therefore, in this paper, we propose a system called HuggingGPT to connect LLMs (i.e., ChatGPT) and ML community (i.e., HuggingFace), which can process inputs from different modalities and solve numerous complex AI tasks. More specifically, for each AI model in HuggingFace, we use its corresponding model description from the library and fuse it into the prompt to establish the connection with ChatGPT. Afterward, in our system, LLMs (i.e., ChatGPT) will act as the brain to determine the answers to the questions of users.

I can already feel the HuggingFace ChatGPT plugin coming. The only problem is it would be very slow and use a bunch of tokens.