Can social media sites use LLMs to make bots that simulate their users?
For example, would it be feasible to augment GPT4 by additionally training it on a particular user's social media comments to create a bot that can simulate that user (but with all of GPT4's capabilities still intact)?
I'd bet that you'd find this idea to be more pervasive on dating sites
Reddit had fake users on day one.