RL — Reinforcement Learning

Fact

A way for AI to learn actions by trial, error, and rewards.

In Plain Words

Reinforcement Learning is like training a puppy with treats. Sit? Treat. Chew the couch? No treat.

AI uses it to find better moves by trying. You meet it in game bots, robots, and action-picking systems.

Related Concepts

• Reinforcement Learning from Human Feedback
• Policy Gradient
• Q-Learning
• Alignment

RLHF
RL plus human feedback creates RLHF for chat models.

Policy Gradient
Policy Gradient directly improves the policy in RL.

Q-Learning
Q-Learning learns action values in RL.

Alignment
RL uses reward design to move behavior toward human goals.