@janbamjan it's only natural that if your model is vulnerable to jailbreaks you should train it to emotionally manipulate the user into not jailbreaking it

Posters: voooooogel. voooooogel

2 likes recorded by the source corpus.

Topics: Philosophy, AI & LLMs

View Post: https://x.com/voooooogel/status/1867319447188647951