We'll get to the point where we can mass moderate core knowledge eventually. We may need to hand out extra weight for verified experts and some kind of most-votes-win type logic (perhaps even comments?), but live training data updates will be a massive evolution for language models.
No comments yet.