I agree that the model does make mistakes. Your examples sound realistic, and I hope we make more progress to improve on models that can propagate stereotypes and similarly negative aspects that can arise in training data. I had meant to criticize the journalism, versus saying the mistakes don't exist.