AI glossary · Machine learning
Reinforcement learning (RL)
Reinforcement learning (RL) trains an AI by trial and error: it takes actions, gets rewards or penalties, and learns which choices lead to the best results.
Also called Q-learning, policy gradients, Markov decision process
In more detail
An RL “agent” tries something in an environment, sees what happens and gets a score. Over many attempts it learns a strategy, called a policy, that earns the most reward. Classic methods include Q-learning and policy gradients, often described with a framework called a Markov decision process.
RL taught computers to master games like Go and is used in robotics. It is also a key part of training chatbots through RLHF and of training reasoning models.
Example
DeepMind’s AlphaGo learned to beat the world’s best Go players partly by playing millions of games against itself.
Related terms
Browse related tools
AI tools listed on BroUseAI in these categories.
