Jailbreaking in verse: how poetry loosens AI’s tongue
Researchers have discovered that styling prompts as poetry can significantly undermine the effectiveness of language models’ safety guardrails.
3 articles
Researchers have discovered that styling prompts as poetry can significantly undermine the effectiveness of language models’ safety guardrails.
The owners of smartphones running Android and iOS sometimes enter into heated discussions as to whether it’s worthwhile to perform the mysterious procedures known as rooting and jailbreaking their devices.
Jailbreaking your mobile device – that is, freeing it from its locked-down status in order to run unauthorized software – used to be for hardcore techies. Not anymore. These days