Comment by quirk
23 days ago
False positives would be way better than worthless. They'd give AI bot mods a fighting chance. If you don't think AI generated writing is detectable, ask Google how they were able to filter spam. Layered signals works.
23 days ago
False positives would be way better than worthless. They'd give AI bot mods a fighting chance. If you don't think AI generated writing is detectable, ask Google how they were able to filter spam. Layered signals works.
> If you don't think AI generated writing is detectable, ask Google how they were able to filter spam.
“Were” is doing a lot of heavy lifting there. Just because Google was able to effectively filter spam in the past doesn't mean they're still able to do so, or that their continued ability to do so has anything to do with LLMs (when you have control over a plurality of the world's inboxen, it's pretty easy to notice when a message is going to an abnormally large number of those inboxen and respond accordingly).