News for hackers. Stuff that matters.

A comment on Anthropic's 'Watermark' Text Adulteration in Claude Is a Perversion of Writing. Parent comment

Re:Anthropic's 'Watermark' Text Adulteration in Claude Is a ... by simonh on Monday August 17, 2026 @01:03PM (#49334235)
But as I explained, the watermark is functionally random PRNG noise overlaid on the token probabilities. It’s not something that can be compensated for because it’s not predictable if you don’t have the seed and PRNG function.
[ Parent | on HN ]
Re:Anthropic's 'Watermark' Text Adulteration in Claude Is a ... by tsimionescu on Monday August 17, 2026 @04:32PM (#49337220)
If it's functionally random PRNG, then how does it differ from any other random sampling? If it's biased PRNG, then the LLM can adapt to the bias, and coincidentally might even benefit from this bias.
[ Parent | on HN ]
Re:Anthropic's 'Watermark' Text Adulteration in Claude Is a ... by simonh on Friday August 21, 2026 @03:24AM (#49384908)
>If it's functionally random PRNG, then how does it differ from any other random sampling?

For practical purposes it isn't.

[ Parent | on HN ]