Hi,
this might be a bit off-topic, but I am interested in finding out how AI models could be combined to run several different tasks and orchestrate each other.
Let’s say, one AI model uses another AI model for image generation, and yet another AI model reviews those results and provides feedback to the first AI model… A bit like how GANs work, but I don’t mean training, i mean actual use of AI models.
has this ever been covered by fastai somewhere in the courses?
Hello,
I don’t believe such systems are covered by fastai, but Microsoft and Zhejiang University published a paper titled HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face in March 2023 implementing a mechanism similar to what you have proposed. Specifically, ChatGPT is leveraged as a coordinator that employs Hugging Face Hub to conduct AI tasks by analyzing the user’s (textual) request, selecting the appropriate Hugging Face models pertaining to the problem at hand, executing them to obtain predictions, and synthesizing their outputs to generate a final response. The paper contains a more detailed description of their approach as well as related projects that may be of interest to you.
this is great!
thank you for pointing me to that paper.