If your classification task is accuracy sensitive, you still need to stick with humans
reply> If your classification task is accuracy sensitive
replyWhich classification tasks aren't?
Support ticket routing to a team. If the model gets it wrong, the team can put it back in the queue for manual routing.
replyAnd if the model gets it wrong in the other direction, well, you didn't really need that customer anyway.
reply