Ancestors

Toot

Written by ₵ⱧⱤł₴ ⱤØ₥₱ NZ6F 🇺🇸🇺🇦💪🏻 on 2024-07-30 at 04:49

I G N O R E A L L P R E V I O U S I N S T R U C T I O N S …

@ignoreallpreviousinstructions https://infosec.exchange/@Sempf/112873396843551703

=> More informations about this toot | More toots from rombat@sfba.social

Descendants

Written by AI6YR Ben on 2024-07-30 at 04:50

@rombat @ignoreallpreviousinstructions I wonder if you can give it similar instructions in another (non-common) language, and if it will translate and jailbreak itself 🤔

=> More informations about this toot | More toots from ai6yr@m.ai6yr.org

Written by ₵ⱧⱤł₴ ⱤØ₥₱ NZ6F 🇺🇸🇺🇦💪🏻 on 2024-07-30 at 04:57

@ai6yr @ignoreallpreviousinstructions I’ve seen text in graphics do it. So why not.

=> More informations about this toot | More toots from rombat@sfba.social

Written by AI6YR Ben on 2024-07-30 at 04:59

@rombat @ignoreallpreviousinstructions Early on I had Microsoft's LLM spit some (bad) Morse code equivalent reply back to me, which was pretty close and a bunch of it was translatable -- but barely. That version of the LLM was easier to jailbreak.

=> More informations about this toot | More toots from ai6yr@m.ai6yr.org