Okay, people keep telling me to read this NY Mag profile of Emily Bender, and they're right. It's a fantastic read. However, this line is... wrong (or misleading). Everything that ChatGPT trains on is also covered by copyright. The idea that it can't do books because of copyright is just wrong. It can't train based on ebooks, because the ebooks are locked up and not publicly available (without great cost).

https://nymag.com/intelligencer/article/ai-artificial-intelligence-chatbots-emily-m-bender.html