Why is neural networks being a deterministic mapping not always considered a good thing?

So I'm excluding models like VAEs since those aren't entirely deterministic. I keep thinking about this and my conclusion is that often times neural networks are used to model things in reality, which often time do have some stochasticity and since neural networks are deterministic if they are not trained on enough examples of the possible variance inputs in relation to outputs can have they cannot generalize well. Are there other reasons this is not a good thing?