KARADAVI

Claude and Constitutional AI

Anthropic's approach to alignment: give the model a written set of principles, then let it critique itself.

Published Invalid Date min read
Anthropic
Editorial overview and documentation for Anthropic.Source: KARADAVI Knowledge Archive

Claude(Model) is the language model family built by Anthropic(Company).

Constitutional AI

Instead of only rating outputs by hand, humans write a set of principles. The model is then asked to critique and rewrite its own responses using those principles.

Positioning

Claude competes directly with ChatGPT(Product) and Gemini(Model) in both product and API markets.

Explore Further

Anthropic(Company)ChatGPT(Product)Claude(Product)Gemini(Product)
Continue exploring the forest
Next trail→Anthropic

Every article leads somewhere. Follow this entity, or search the whole forest.