Researchers at OpenAI have developed an algorithm that uses reinforcement learning to learn high-level actions useful for solving various tasks. This approach allows agents to solve complex problems requiring thousands of time steps more efficiently. The algorithm, applied to navigation tasks, discovers a set of general actions such as walking and crawling in different directions, enabling the agent to quickly master new navigation tasks. --- Why it matters: This matters because it could enable AI systems to adapt more easily to new situations, reducing the need for extensive retraining or manual intervention. This is particularly relevant for applications where environments are constantly changing or when agents must learn from limited data. Source: https://openai.com/index/learning-a-hierarchy