It looks like the results in this work can indeed have some impact on the real world and legislation, which is more than what most papers at Nature and elsewher

It looks like the results in this work can indeed have some impact on the real world and legislation, which is more than what most papers at Nature and elsewhere accomplish. In 2027, when 90% of the Internet is AI slop, we won't even be able to train new foundational models unless we maintain the sanity of the Internet... And we need to start doing something about it. Kudos to the authors.