> Filtering shitty content is easier than creating it with a properly constructed LLM system Can you explain this? It feels completely wrong - even OpenAI, who

> Filtering shitty content is easier than creating it with a properly constructed LLM system Can you explain this? It feels completely wrong - even OpenAI, who probably have invested the most, can't filter out all "shitty content". An LLM can on the other hand create "shitty content" incredibly easy - even if the creators try to stop it! So how is filtering easier than creating?