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