If the model's responses are chosen by a human, then it can be thought of as a kind of reinforcement learning, the model explores the space of possibilities by

If the model's responses are chosen by a human, then it can be thought of as a kind of reinforcement learning, the model explores the space of possibilities by generating samples, the human is the reward model.