Those interacting with large AI language models note that these AIs possess distinct 'personalities.' Experts suggest that this variability is determined by how the models were specifically designed to interact with human users.