Reinforcement learning with human comments (RLHF), by which human consumers evaluate the precision or relevance of model outputs so the product can increase by itself. This can be as simple as owning persons kind or communicate back again corrections to a chatbot or Digital assistant. Unsupervise