I remember "arguing" (just for fun) with a language model once that instead of trying to make them appear safe and trustworthy, the "safest" and most ethical action would be complete transparency and for the chatbot to always tell people "Don't listen to AI you retard, do your own research". In the end the chatbot said I was right.
If you believe AI can be safe, secure and trustworthy, you already fucked up.
I remember "arguing" (just for fun) with a language model once that instead of trying to make them appear safe and trustworthy, the "safest" and most ethical action would be complete transparency and for the chatbot to always tell people "Don't listen to AI you retard, do your own research". In the end the chatbot said I was right.