I was being vague, but I meant those questions in the context of an ML model.

ML models are completely inference-based. They have zero symbolic definitions. They don't categorize subjects at all.

Because of that, you can't tell a model how many X to draw.