A study of more than 70 large language models found similar answers to brainstorming and creative writing prompts; originality gets stripped away and the process favors safe, consensus-based responses ...