Researchers at Northwestern University have developed an algorithm called Maximum Diffusion Reinforcement Learning (MaxDiff RL) that allows robots to learn new tasks more quickly. Unlike traditional reinforcement-learning algorithms that require hundreds of thousands of attempts to master a task, MaxDiff RL focuses on training robots to explore a wide range of experiences. The algorithm encourages robots to be randomly adventurous and maximize the diversity of state changes rather than actions. In tests, robots powered by MaxDiff RL outperformed other state-of-the-art reinforcement learning algorithms in a simulated swimming task. The algorithm immediately adapted its learned behaviors to the new task, while other algorithms required multiple resets to figure out how to move. However, the researchers note that this does not mean MaxDiff RL can be directly applied to real-world scenarios like self-driving cars. [ae47d92c]