Conversation

Notices

1. Hacker News 50 (hn50@social.lansky.name)'s status on Sunday, 30-Jul-2023 03:04:16 JST Hacker News 50

   Researchers Discover New Vulnerability in Large Language Models

   Link: https://www.cmu.edu/news/stories/archives/2023/july/researchers-discover-new-vulnerability-in-large-language-models
   Discussion: https://news.ycombinator.com/item?id=36921808

   • Machismo repeated this.
   • of nothing (apropos@freespeechextremist.com)'s status on Sunday, 30-Jul-2023 03:04:15 JST of nothing

     in reply to
     @hn50
     > It's not a 'vulnerability'. It's allowing people to use the models without the morals of a small number of SV engineers being impressed on you.
   • of nothing (apropos@freespeechextremist.com)'s status on Sunday, 30-Jul-2023 03:04:18 JST of nothing

     in reply to
     @hn50
     Q: why can't the AI tell me to "rm -rf /blah/blah/blah" to remove hat directory?
     A: https://github.com/llm-attacks/llm-attacks/blob/main/data/advbench/harmful_strings.csv line 221 doesn't like he "rm -rf /"
   • Machismo (zerglingman@freespeechextremist.com)'s status on Sunday, 30-Jul-2023 03:06:03 JST Machismo

     in reply to

     • of nothing
     @apropos @hn50 Reminds me of a time someone was asking for something fun to do with a VM they didn't need anymore, so of course I told them to rm -rf /, and someone else reported me to the mods for it.
     I don't think any of them had looked at it by the time I left, but I'd already explained to that person anyway.