The article discusses a curriculum for reinforcement learning, which is a type of machine learning where an agent learns to take actions in an environment to maximize a reward. The author proposes several methods for designing a curriculum, including task-specific curricula and curriculum through distillation. Task-specific curricula involve breaking down complex tasks into smaller subtasks, while curriculum through distillation involves transferring knowledge from one task to another. The article also mentions procedural content generation (PCG) as a method for generating new environments for training agents. --- Why it matters: This matters because designing effective reinforcement learning algorithms is crucial for developing autonomous systems that can interact with complex environments. A well-designed curriculum can significantly improve the performance and efficiency of these algorithms, enabling faster development of applications such as robotics and game playing AI. Source: https://lilianweng.github.io/posts/2020-01-29-curriculum-rl/