We are already headed to this point but just imagine: In the future you get some answer from AI and you want to see if it’s true. Where would you check? Ask another AI or the same AI if it’s right? I can just see a potential future where AI could pretty much tell us anything and we would have way to know if it’s right or not.
AI-assisted search, sans summaries.
Don’t need it to try its best to be up to (but not always including) 100% correct—can just use its capabilities to ease the manual process.
Then can work on “what about when it pulls accurately-verbatim citations from absolutely-awful sources” :)