So one of the major problems with #ChatGPT is that it's deliberately opaque about the basis for the claims it makes. What happens if you explicitly ask it to cite its sources?
Turns out the answer is - it will just spit out citations for non-existent papers and falsely attribute scientific discoveries to people who didn't make them!
(local differentially private mechanisms have existed since at least the 60s, and have been studied by that name since at least 2008: https://arxiv.org/pdf/0803.0924.pdf)