LLMs can get "brain rot"

After reading this, I just felt like everyone already knows the data is a mess, but no one really cares. We feed the models a bunch of junk, then act surprised when they start getting dumber. Honestly, did we even need a study to figure that out?