Another day, another preprint paper shocked that it’s trivial to make a chatbot spew out undesirable and horrible content. [arXiv] How do you break LLM security with “prompt injection”?…
awful is great and i’m glad that it’s a thing, but there’s entire world beyond it, and to go there, i curate a shitlist (now less intensely than previously). this can tell you to avoid things like any pleroma instance, for example (and if it was up to me, i’d defederate by default from them)
well for one it looks like that wasn’t one off and more of a pattern. i remember this one https://lemmy.world/post/25606000/15124978 and i’ve heard of this one https://lemmy.cafe/post/13433291/9941167 also in modlog you can see that that account was mass-banned from ml subs just five days ago so it’s not some ancient incident
awful is great and i’m glad that it’s a thing, but there’s entire world beyond it, and to go there, i curate a shitlist (now less intensely than previously). this can tell you to avoid things like any pleroma instance, for example (and if it was up to me, i’d defederate by default from them)
right right, but we only see these bozos when they show up locally
mind you there’s been more than one illustrious poster I’ve banned preemptively
but frankly life is too short in most cases