3 ms·
That genuinely surprises me, as the sorts or errors made or at least non-idiomatic English used by non-native speakers are not something you typically see from
by wizofaus 3y ago
That genuinely surprises me, as the sorts or errors made or at least non-idiomatic English used by non-native speakers are not something you typically see from LLM output. Though I don't doubt you could get them to generate text "as though" it were written by a non-native speaker.
- powersnail 3y agoIf the detector biases towards the detection of certain qualities of native speakers' writing, using the lack of which as a marker for GPT generated content, it would then logically place an unfair bias on non-native speakers.