Did OpenAI and others pay for the training data from Stack Overflow, Twitter, Reddit, Github etc. Or any other source produced by mankind?