this is from a thread on anthropic marking #ai_slop in #writing or #writingcommunity and i realized it really should be posted as a standalone: There’s two forms fear around marking AI in writing can take. Form #1 is “it will identify my writing as AI.” As general solution, this is correct. But this particular case, it’s a misunderstanding. There is no “ground truth” to be preserved. The cryptographic approach means that the shift is intentionally buried in the randomness inherent to the “stochastic parrot.” And a human is never going to be able to implement the distortions of any fixed key accidentally *or* intentionally. Form #2 is the interesting one. One could also be scared that when you have LLMs write boilerplate (one of the three ways I imagine them being useful; the other two are brainstorming and Orwellian satire), the whole will be marked as LLM written with a high probability. And I’m OK with that. When someone is clear about their use of the technology, the fear about being “caught up” *goes away*. And that gets to the real problem. People look at the tech, see that it helps them get their point across, and then get scared that people will ignore their ideas because their words get flagged as AI. But think about it. If you’re doing something where the words themselves carry little meaning, why do you care? Communication depends on what is perceived, not how or even what was written. But if you’re doing something that represents *you?* That’s exactly when the LLM is stealing your voice. If you have to disclose, you don’t feel like you’re representing yourself in the text. Good. You’re not. Any more when you grab a page from Wikipedia and try to pass it off as your own. Would you do *that* and not think it was plagiarized? And that gets to my last point. This is not theoretical for me. I write quite a bit. The vast majority is not “AI assisted” in strict senses. I talk to the LLM, close it, write words. But *even doing that* I see LLM patterns in my text. I wrote it this way: “my brain is polluted.” So I disclose. And the result has been literal attacks from the community. But someone has to try to bring sense into this discussion. As it is, nobody is even able to *see* the truth, let alone act on it.