All of this! We definitely have to stop policing and outing people. If you don't like a book, don't read it. I saw a post the other day saying you can tell something is written by AI because it has illogical errors in the narrative that don't make sense. Well, I was just re-reading my work in progress that came from my human brain and there is a character in chapter 13 engaging with people who very clearly went somewhere completely else in chapter 12.
There's a big difference in asking a model to output an edited version of the writing, and using AI or other tools like Grammarly to suggest edits. The former will always imprint statistical biases and signature patterns that are almost impossible to remove in standard LLM sampling (I'm working on alternative samplers that might diminish it).
All of this! We definitely have to stop policing and outing people. If you don't like a book, don't read it. I saw a post the other day saying you can tell something is written by AI because it has illogical errors in the narrative that don't make sense. Well, I was just re-reading my work in progress that came from my human brain and there is a character in chapter 13 engaging with people who very clearly went somewhere completely else in chapter 12.
Hahaha omg right? My WIPs are a mess! How HUMAN!!!
There's a big difference in asking a model to output an edited version of the writing, and using AI or other tools like Grammarly to suggest edits. The former will always imprint statistical biases and signature patterns that are almost impossible to remove in standard LLM sampling (I'm working on alternative samplers that might diminish it).