Is there any way to detect, predict the likelihood of a text being generated by GPT-3. It seems like there is something like that for GPT-2 here: gpt-2-output-dataset/detector at master · openai/gpt-2-output-dataset · GitHub
but I haven’t been able to find anything similar for GPT-3.
As the text produced by GPT3 (using for example the playground) are not detectable by existing plagiarism detection tools, I foresee problems in the education field.
It is my understanding that unlike GPT2, GPT3 is not sharing it’s models in a way that would allow a detector tool such as the one above. In that case would it be possible / desirable to expose such a mechanism via an API?