“One thing I like about this approach is that, because it never goes inside the neural net and tries to change anything, but just places a sort of wrapper over the neural net.”
– Computer scientist Scott Aaronson, discussing cryptographic watermarks he’s developing for
OpenAI’s GPT language model. “We want there to be an otherwise unnoticeable secret signal in its choices of words,” he says of encoding specific vocabulary and syntax patterns that will make AI-generated texts instantly detectable, protecting against both plagiarism and propaganda.
