GPT detectors are biased against non-native English writers

Re:GPT detectors are biased against non-native English writers by golergka on Sunday May 21, 2023 @12:42PM (#36022627)

> What exactly makes anyone think that they can detect an LLM that is outputting text?

Watermarking.

On each word of the output, you randomly split all possible words into two groups and only generate output using one of them. If you get a text of 1000 words that exactly follow the secret sequence of the groups, you can be sure that it's generated by this LLM with one in 2^1000 chance of error.

[ Parent | on HN ]

Re:GPT detectors are biased against non-native English writers by dragonwriter on Sunday May 21, 2023 @12:52PM (#36022709)

> Watermarking.

Watermarking only works if every LLM system available does it and it is impractical for third parties to spin up and use systems that don’t, otherwise it only detects content that opts in to being detected.

[ Parent | on HN ]

Re:GPT detectors are biased against non-native English writers by Der_Einzige on Sunday May 21, 2023 @09:04PM (#36026252)

This technique is trivially fooled by anything which messes with the Logits, which is what everyone is now doing in the non chatGPT world.

[ Parent | on HN ]