Foundation models on federated setting

Prompt engineering is a major way of using foundation models, e.g. ChatGPT. However, privacy-preserving on prompt learning would be a new risk of using foundation models. For example, a user can create some prompt using her/his own emails or private documents that should not be shared out of their own computers. For this case, federated learning will be a new mechanism to be integrated with prompt learning to provide a privacy-preserved use of foundation models.