OpenAI claimed it’s “impossible” to build good AI models without using copyrighted data. An “ethically created” large language model and a giant AI dataset of public domain text suggest ...