The authors speculate about the implications of this new detection method for detecting AI-generated text.
They see it as becoming more important as machine-learning techniques advance rapidly.
Specifically, they focus on the detection of text in terms of true positive rate and false positive rate (FPR) , which adds an orthogonal perspective to the current debate on detectability.
Also, the detection method provides the capability for model sourcing, enabling the identification of the specific language model used for text generation.
The authors discuss the theoretical analysis that focuses on the possibility of AI-generated text having two false positive rates, TPR and PR, which add anorthogonistically to the currently debated issue of detectability vs. false effective rate.