OpenAI Wants AI to Help Humans Train AI

ChatGPT's success can be attributed to the human trainers who trained the AI model to differentiate good and bad responses. OpenAI initiated the use of reinforcement learning with human feedback in the development of ChatGPT. In this technique an input is taken from human testers to fine-tune the AI and make the responses more coherent. However, human data can be inconsistent, how can AI help?