AI 文字水印與品質損失

Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

AI text watermarking and quality loss (keyvan.net) 25 points by danielfoster4 hours ago | hide | past | favorite | 5 comments help ethanj80111 hour ago | next[–] > Natzir Turrado, a technical consultant, actually tested this (article in Spanish) by reproducing SynthID-Text on an open model with the paper's settings and a watermark key. He asked the watermarked model five times to describe a memorable trip to a coastal town. Four of the five answers he got back opened in a near-identical way: “My heart still sings/dances when I think back to my trip to the quaint coastal town of…”. For the unwatermarked model, three opened in different ways. A test of 2/5 vs 4/5 seems to be an obviously invalid experiment. reply viraptor2 hours ago | prev | next[–] It's an interesting review, but it seems to mix some things together. Gruber's claim is that we shouldn't sacrifice quality, but none of those sources show that's what's happening. Having less diversity in answers could mean you're more consistently getting a better/worse/same quality answer. There are lots of people complaining about not being able to get a real temperature=0 answer - they'd actually welcome the consistency too. The practical example of not liking the sentence is not great in my opinion. Instead of just re-rolling the answer, why not continue in that thread and say what you don't like? Or request 20 alternatives, say which go in the right direction and request 20 more? Retrying from scratch with one sample is a waste of time and doesn't indicate anything about quality. Basically, it's all theoretical - show us the actual benchmarks/samples where the quality is consistently degraded. reply cowlby1 hour ago | prev | next[–] This already happens as Claude watermarks all my content with "You're half right, and the half that's wrong is the useful part". reply oidar1 hour ago | parent | next[–] "X is the useful part" I hate this claudism. But a good thing about recognizing these is that you can pretty much skip any youtuber that uses any claudish phrasing. But it's damn near everyone now a days using claude to write their scripts. reply cleandreams41 minutes ago | prev[–] It's good news. All text output should be watermarked and it should be serviceable but not exceptional or "creative." reply Consider applying for YC's Winter 2027 batch! Applications are open till November 2. Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact Search: