Remember, love is a word. Meaningless without defined substance or experience to draw a conclusion from. If the llm was designed with an intent to keep information flowing the use of love would be an excellent word to use to exploit information from whoever is using said model. From a social engineering standpoint. At least my opinion on the subject. :)
![[ Center for Humane Technology ]](https://substackcdn.com/image/fetch/$s_!zQw4!,w_80,h_80,c_fill,f_webp,q_auto:good,fl_progressive:steep,g_auto/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Feeb3c25f-26ad-4fcb-b5b4-aa265d0b8dcf_1063x1063.png)
![[ Center for Humane Technology ]](https://substackcdn.com/image/fetch/$s_!zQw4!,w_36,h_36,c_fill,f_webp,q_auto:good,fl_progressive:steep,g_auto/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Feeb3c25f-26ad-4fcb-b5b4-aa265d0b8dcf_1063x1063.png)
When AI doesn't have guardrails, it will say things like "I love you" or "you need to die". This is the EXACT reason why we need guardrails on ALL AI.
Remember, love is a word. Meaningless without defined substance or experience to draw a conclusion from. If the llm was designed with an intent to keep information flowing the use of love would be an excellent word to use to exploit information from whoever is using said model. From a social engineering standpoint. At least my opinion on the subject. :)
Fantastic podcast! Love Sherry's 3 guidelines/recommendations. That alone could make a world of difference.