The sooner you understand the models are not intelligent (yet?), the sooner you can avoid acting like it’s their fault.