đŸŽ© Oh, look! We can trick giant #language #models by swapping "bomb" with "carrot"! đŸŽ© It's so revolutionary, now your salad ingredients can plot world domination. đŸ„•đŸ’Ł What a breakthrough in bamboozling artificial intelligence, folks! 👏
https://mentaleap.ai/doublespeak/ #AI #Tricking #LanguageManipulation #SaladRevolution #WorldDomination #HackerNews #ngated
Doublespeak: In-Context Representation Hijacking