3 ms·
OpenAI should release a classifier that detects their own AI-generated text. They could do this easily by just using steganography to hide some information in
by andrewmutz 4y ago
OpenAI should release a classifier that detects their own AI-generated text. They could do this easily by just using steganography to hide some information in all text that they generate, and then build the classifier to look for it.
Sure, it's less useful than a classifier that can detect any AI generated text, but it would be a nice tool for contexts where AI generated text can be abused (like the classroom) in the short term.
- m3affan 4y agoThere is work on hidden signatures in generated text, invisible to humans. Only way to move forward.
- thewataccount 4y agoThe problem with this will be the method to detect the signature would reveal how to hide the signature though right? Obviously not an issue if everyone uses a single API for it - but if this ends up like Stable Diffusion were anyone can run it locally then I don't think it's possible no?
- bnug 4y agoI'd think people would migrate to just re-typing whatever was generated and change some wording along the way to prevent detection.
- fodi 4y agoOr another AI could just do that.
- ineedtocall 4y agoOr they could just save/hash results and get rid of the classifier all together.
- rcme 4y agoYea, they could provide a fingerprinting algorithm and a database of every fingerprint they've generated. However, it wouldn't help you identify false-positives.
- sebzim4500 4y agoScott Aaronson talks about something like that being done at OpenAI in this post https://scottaaronson.blog/?p=6823 https://scottaaronson.blog/?p=6823