Introduction
How do you teach a dog to sit? You don't show it thousands of labelled examples of "sitting". You let it try things, and when it sits, you give it a treat. Over time it learns which actions lead to treats.
Reinforcement learning (RL) works the same way. An agent interacts with an environment, takes actions, and receives rewards. Nobody tells it the correct action; it must discover by trial and error which behaviour earns the most reward over time.
RL has produced some of AI's most famous results: programs that beat world champions at Go and complex video games, robots that learn to walk, systems that cool data centres more efficiently, and the fine-tuning step that makes language models more helpful (topic 1.5). Deep reinforcement learning uses neural networks to handle complex inputs like images.