Constitutional AI

Fact

A training method where AI uses rules to critique and fix its own answers.

In Plain Words

Constitutional AI is like a family rule sheet on the fridge. Before the AI blurts out a wild answer, it checks the rules and fixes itself.

You meet it in safety training for chatbots. It can reduce harmful replies and need fewer human graders.

Related Concepts

• Alignment
• Reinforcement Learning from Human Feedback
• Post-training

Alignment
Constitutional AI writes values into the model's behavior.

RLHF
Constitutional AI uses rules and AI feedback, so it needs less human feedback.

Post-training
Constitutional AI usually happens after pre-training, during model tuning.