Yep that's why I said let's not even consider newspapers. Many of those have been using AI generated/content-mill/sponsored content for years and years.

Also why I acknowledged journal articles can be worth less than the paper they're printed on. Even if you were to select for reputable, high impact journals, those also often experience scandals, retractions, potential data fabrication, etc.

But then there are textbooks and technical publications, also being published in the hundreds of thousands globally each year.

The fact is that with billions of human beings on the planet, and media increasingly being digitized by default, and AI-content detection, I don't see how we could possibly run out of new content to grow LLMs...