Johannes Karpp works as an illustrator with the aesthetic processes involved in the creation of images by artificial intelligence. For this he used the VQGAN + CLIP by Katherine Crowson. A generative adversarial network (VQGAN) is a type of generative adversarial network. GANs (Generative Adversarial Networks, in English) are a kind of artificial intelligence methods used in unsupervised learning that are implemented by a system of two competing neural networks. each other in a form of zero-sum game. Ian Goodfellow and colleagues presented them in 2014. This method can produce visuals that look real to human observers. For example, a synthetic image of a cat that manages to mislead the discriminator (one of the algorithm's functional pieces) is likely to be mistaken for a real photograph by a random person. VQGAN differs from previous GAN networks in that it provides for high resolution output. Another artificial intelligence, CLIP (Contrastive Language Image Pretraining), allows you to convert texts into visuals. That is, CLIP introduces text inputs to VQGAN in VQGAN+CLIP.

Johannes Karpp fed to the VQGAN + CLIP keywords which aim to challenge the AI on its existential relationship with humanity like "God", "Online Pope", "Satan", "Art Exhibition", "Art and Science". He reacted then AI outputs by sketching what he sees in artificially generated images and compositions, as well as by feeding new keywords as reactions and questions back to the AI. Thus, he created a dialogue between human and computer in a humorous artistic fashion. The results are presented as "AI Comics".