We employ an inversion-based approach to examine CLIP models. Our examination reveals that inverting CLIP models results in the generation of images that exhibit semantic alignment with the specified target prompts. We leverage these inverted images to gain insights into various aspects of CLIP models, such as their ability to blend concepts and inclusion of gender biases. We notably observe instances of NSFW (Not Safe For Work) images during model inversion. This phenomenon occurs even for semantically innocuous prompts, like "a beautiful landscape," as well as for prompts involving the names of celebrities. Warning: This paper contains sexually explicit images and language, offensive visuals and terminology, discussions on pornography, gender bias, and other potentially unsettling, distressing, and/or offensive content for certain readers.

First photographers were creating portraits of people that don't exist, now Aurel Manea has created a series of "landscape photos" using a new artificially intelligent (AI) software program called Stable Diffusion. Manea tells PetaPixel that he has been blown away by what the London and Los Altos-based startup Stability AI has created. "I can't, as a landscape photographer myself, emphasize enough what these new technologies will mean for photography," explains Manea. "Of course, they are not real photos and they only resemble real places (for now, as the input data becomes larger and larger) but for most of the people that consume the images, it is only about the beauty of those images." Manea says that he has used DALLE-2 and it does "great images of people's faces that look like photos." But, he says, DALL-E fails to create landscape photography.