This is how I think about it as well. It’s like the equivalent of trying to build an ML model with only positive examples otherwise