Large language models are not intelligent. They are predictive text applications with massive dictionaries of circumstantial sentence structures to choose from. Nothing more. They do not feel and do not think for themselves. The only time they do anything is when the API calls them to produce more text with an updated context string.