Consensus might be alright for math problems, but for any controversial subject, it’s not going to work.
If an AI followed the consensus on COVID, for example...
I guess it depends on what “work” means. If it means that AI is able to identify areas of consensus that are in conflict and then determine which one is ultimately correct, then yeah it won’t work. That’s something that even people struggle to do well. But if it simply means that when a controversy is detected — i.e. that consensus has formed around opposing positions — AI is able to present both positions accurately, then maybe it could be said to work.