3 ms·
I'm extremely skeptical of these claims. Especially when we're dealing with careful prompting to adjust tone/style.
by alfalfasprout 1y ago
I'm extremely skeptical of these claims. Especially when we're dealing with careful prompting to adjust tone/style.
- zingababba 1y agoMmmm yes, I probably will never be able to find it again but someone recently tested a lot of these out and found you could bypass them easily by changing a few words around.
- Mkengin 1y agoOr use RL to beat any AI detectors: https://reddit.com/r/LocalLLaMA/comments/1lnrd1t/you_can_just_rl_a_model_to_beat_any_ai_detectors/ https://reddit.com/r/LocalLLaMA/comments/1lnrd1t/you_can_jus...
- haffi112 1y agoEven if it was close to being near perfect, that is still not enough due to the negative impact of false positive detections on students.