>it assumes that content on the internet is not moderated and that the training dataset will never evolve to take rating into consideration Worse than this, si

>it assumes that content on the internet is not moderated and that the training dataset will never evolve to take rating into consideration Worse than this, since the training set already takes rating into consideration. For example the reddit dataset includes only articles with more than 3 upvotes.