Is this problem solveable? A model trained to optimize for what happens next in a sentence is not ideal for interaction because it just emulates bad human beha

Is this problem solveable? A model trained to optimize for what happens next in a sentence is not ideal for interaction because it just emulates bad human behavior. Combinations of optimization metrics, filters, and adversarial models will be very interesting in the near future.