After a rise in quirky creature metaphors in ChatGPT responses, OpenAI has fixed the issue by removing the training ...