Text-to-image AI models can be tricked into generating disturbing images

Researchers have developed an algorithm that creates nonsense words Stable Diffusion and DALL-E 2 recognize as commands to produce pictures showing violence and nudity.

Related Keywords

Israel ,Alex Polyakov ,Yinzhi Cao ,Neil Zhenqiang Gong ,Zico Kolter ,Duke University ,Carnegie Mellon University ,Johns Hopkins University ,Carnegie Mellon ,Text To Image Ai Models Can Be Tricked Into Generating Disturbing Images ,