Using a complex GPT-4o prompt, they sought to pull out tweets that focused on “superficial topics”

Wait a moment… They asked an LLM, to tell them what was “junk”, and another LLM, trained on what an LLM marked as junk, turned out to be a junk LLM?

It talks about model collapse, but this smells like research collapse.

deleted by creator

The problem comes from people responding to them, in communities that are text-first. You can flag a person as a troll… but unless you ban them, then eventually the discussions spill over.

Overwatch doesn’t have that same issue, because text is not at its core; you can mute people, and keep playing the game. Game-disrupting players though, still get banned.