We’ve trained a model called CatGPT which interacts in a conversational way like a cat. The dialogue format makes it possible for CatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.
