I think it’s a great idea for OpenAI to train its models on StackOverflow’s massive trove of obsolete technical information.