Constitutional AI
Domain: LLM Concepts
🎯 Motivation
An alignment method where the model follows a set of principles (a constitution) to self-correct its responses, reducing the need for human feedback.
📋 Concrete Example
Constitution: ‘Be helpful, harmless, honest.’ Model generates a response, checks it against the constitution, revises if needed. ‘I should not provide hacking instructions - revising to explain security instead.’
🔗 Analogy
A student with a personal code of conduct - before submitting an answer, they check: ‘Is this honest? Is this helpful? Is this appropriate?’ They self-correct based on their principles (constitution).