Topic

AI text watermarking

4 mentions in 2 episodes

A method of making AI generated text identifiable that the hosts discuss alongside its detection limits.

Mentions

758: Perverts at Large!

  • 47:43

    Text watermarking can subtly alter the statistical choices a model makes.

  • 48:00

    Paraphrasing, translating text into another language and back, or mixing AI output with human writing can weaken detection.

660: The Bendy Piece

  • 21:37

    research into text watermarking for their AI

  • 22:15

    doesn't perform as well when faced with certain forms of tampering