One thing to note from the USPTO doc is that it describes the reproduction of substantial parts of the source corpus as an error condition. This bears on part o

One thing to note from the USPTO doc is that it describes the reproduction of substantial parts of the source corpus as an error condition. This bears on part of the Author's Guild complaint, that ChatGPT was at one point able to reproduce chunks of copyrighted text. OpenAI would like to persuade USPTO that this would be an erroneous output of ChatGPT. This seems like a persuasive position, but not necessarily dispositive.