I think it'd be really nice for there to be an effort within GPT to have responses that don't mimic the bias of information found online and instead draw its own conclusions based evidence, and the ability to scrutinize different types of evidence.