Viewing a single comment thread. View all comments

wobrob101 t1_j3ee2lj wrote

If the human errors are random then yes, if there is a consistent pattern that allows the model to learn the "correct" errors then that can cause problems

1