This is a really tough problem. I don't think ignorance is the answer, but it's also difficult to set aside things that seam legitimate and go down a rabbit hole of reinventing something on a hunch. I guess the saving grace is that it's impossible to know enough about such a wide swathe that it's often a problem. With large models that conceivably can encode the collective knowledge, though, we have to be vigilant about creating an orthodoxy that ultimately constrains us.