AFAIK GPT-3 was not trained only on public domain data?
Are there cases of GPT reciting stories? Most content I've seen from it has seemed original enough that I couldn't find a source by searching parts.
AFAIK GPT-3 was not trained only on public domain data?