constitutional AI

Definitions

1. A training methodology developed by Anthropic where an AI system is guided by a set of written principles (a 'constitution') rather than relying solely on human feedback for every decision. The model critiques and revises its own outputs according to these principles, reducing the need for extensive human labeling.

   In plain English: Teaching an AI to follow a written set of rules about being helpful and safe, so it can check its own work instead of needing a human to review everything.

   Example: Their constitutional AI approach uses a set of 15 principles including helpfulness, harmlessness, and honesty to self-correct before the response reaches the user.

Related Terms

• AGI
• Alignment
• Bias
• AI Safety
• AI alignment
• guardrails
• AI Alignment Problem
• Safety Alignment