One-Sentence Definition
Reinforcement learning (RL) is a type of machine learning where an agent learns to make decisions by taking actions in an environment and receiving rewards or penalties based on the outcomes.
How It Works
In reinforcement learning, there is no labeled dataset. Instead, an agent interacts with an environment over many episodes. At each step, the agent observes the current state, chooses an action, receives a reward signal, and transitions to a new state. The goal is to learn a policy -- a mapping from states to actions -- that maximizes cumulative reward over time.
The challenge is credit assignment: when a chess game ends in a loss 40 moves later, which early move was the mistake? RL algorithms solve this through techniques like temporal difference learning and Monte Carlo methods that propagate future rewards backward to earlier decisions. Deep reinforcement learning combines these ideas with neural networks, letting agents handle high-dimensional inputs like raw game pixels or robotic sensor data.
Famous milestones include DeepMind's AlphaGo (2016), which beat the world Go champion using a combination of deep RL and tree search, and OpenAI Five (2019), which defeated professional Dota 2 teams. More recently, reinforcement learning from human feedback (RLHF) became a critical ingredient in making large language models like ChatGPT and Claude more helpful and less harmful. In RLHF, human raters rank model outputs, a reward model is trained on those rankings, and the LLM is fine-tuned to maximize the reward model's score.
Why It Matters
RL is the go-to paradigm when you need an AI to make sequential decisions under uncertainty. It drives autonomous vehicle planning, robotic manipulation, data-center cooling optimization (Google used RL to cut cooling energy by 40%), and game AI. Its role in LLM alignment -- shaping models to be safe and useful through RLHF and its successors like DPO (direct preference optimization) -- makes it one of the most consequential techniques in current AI development.
Key Takeaway
Reinforcement learning teaches agents through rewards and penalties rather than labeled examples, and it is essential both for decision-making AI and for aligning large language models with human preferences.
Part of the AI Weekly Glossary.