Adding a personality to ChatGPT is accomplished during its training process. The model is trained on a diverse range of internet text. However, it doesn’t know specific documents or sources it was trained on and doesn’t have a specific personality. The ‘persona’ it exhibits in a conversation is a fabrication based on the instructions given to it.
If you want the model to exhibit a certain personality in a conversation, you can instruct it as such. For instance, you could tell it to talk like a particular character, person, or personality type. But it doesn’t ‘know’ what it’s like to be that entity – it’s more akin to an actor following a script.
One research direction OpenAI hints at is to allow users to easily customize ChatGPT’s behaviour to suit individual needs, which might make it possible to add more specific personality traits to the chatbot in the future. However, there would have to be some bounds on system behavior to prevent misuse. OpenAI is still in the process of defining guidelines about system behavior and deployment policies.
