I think that depends on how you calibrate the ship of Theseus: If someone uses an LLM to "proofread" and it suggests one word instead of another, does it stop being the author's work? Two words? Three? Switching the order of two sentences?
I think the key concern here is that a statistical "watermarking" system could insert a positive signal even when the gradual changes are--to any human watching the editing process--not things we would consider disqualifying.
It may just be a misreading but say you asked for a review of the text and it silently water marked it as its own. Guess source control to review changes is the solution.
> My writing is my work, and Anthropic’s current strategy is aggressively writer-hostile.
If your writing is indeed your work then watermarking should not affect you.
I think that depends on how you calibrate the ship of Theseus: If someone uses an LLM to "proofread" and it suggests one word instead of another, does it stop being the author's work? Two words? Three? Switching the order of two sentences?
I think the key concern here is that a statistical "watermarking" system could insert a positive signal even when the gradual changes are--to any human watching the editing process--not things we would consider disqualifying.
Rands argument is that the following actions will also water mark their hand written text.
Synthesis Proofreading Translation Summarization Critique
It may just be a misreading but say you asked for a review of the text and it silently water marked it as its own. Guess source control to review changes is the solution.
>What this means: if I ask Claude to critique a piece and I happen to cut and paste part of its reply into the text, then that could be marked.
...that seems correct to me? Why wouldn't it be marked? If you take the model's suggestions and apply them, those should absolutely be marked?