Comment by InvisibleUp

If the output of this is even somewhat coherent, it would disprove the argument that mass amounts of copyrighted works are required to train an LLM. Unfortunately that does not appear to be the case here.