OpenAI's big idea to increase the safety of its tech is to have AI models police each otherOpenAI is experimenting with AI models explaining their thought processes to each other for transparency.