RL-Teacher is an open-source tool for training AIs using occasional human feedback instead of pre-defined reward functions. This approach aims to improve the safety and efficiency of AI systems, particularly in situations where specifying rewards is challenging. --- Why it matters: This matters to researchers because it provides a practical solution for training AIs with complex or hard-to-specify objectives, which can be a major hurdle in developing safe and effective reinforcement learning systems. Source: https://openai.com/index/gathering-human-feedback