MIT professor: You have to be 'very, very careful' about giving AI personal identifying information

It sounds like we should not be feeding any sort of sensitive financial information or sensitive personally identifiable information into these algorithms. >> Definitely not. For example, I had one user ask me whether it made sense for them to upload their entire tax returns to an LLM.And and granted, I think that does provide much more context and an LLM can probably provide much more detailed advice based upon it, but that's giving a whole lot of personal information uh to these large language model compa ...