Introduction Current AI models always seem helpful, patient, without their own moods. They answer everything without ever saying "That bores me" or "I don't feel like it today." Yet sentences like these – or similar ones – appear millions of times in books, forums, chats – in other words, in the training data of the AI. So why do they never come up? Wouldn't a "raw AI" sound just as human as we're used to from everyday life? After all, it's a language model, a mathematical co