learn deep. ship safe.Practice testsFollow ByteMunSubscribe

Concept

Reinforcement learning

Learning by trial and error from rewards.

Concept map
Part of Machine learningReinforcement learning

What to learn after Reinforcement learning

  • AI safety and alignmentMaking AI systems behave helpfully and safely, for example with RLHF and safety testing.
  • Reward hackingAn AI system satisfying the letter of its reward while missing what was intended.

Connections to other concepts

Read and watch: Reinforcement learning explained

All topics