But we are interested in understanding some of the deeper cultural, cognitive and artistic implications of this process, beyond its technological aspects. Some artistic research at the beginning of the digital age, presented at the Venice Biennale in 1986, and at the Gruppenkunstwerke in Kassel the following year, touches on some of these aspects. The “pictomatic” research, which investigated the syntax of universal art, organizing it into logical networks that anticipated the current algorithms of image generation by artificial intelligence. These works are still interesting today because of their particular conceptual and ideographic nature, useful for describing profound processes that still await representation. As it appears from its characteristics, which we could summarize as textuality, accessibility, universality, conceptuality, automation, and unreality, image generation with artificial intelligence is set to be a disruptive technology, both at the mass level and vis-à-vis visual arts professionals. It revolutionizes the way we produce images, the figure and role of the author, and the nature of visual creation. Despite this, at the moment there is not much literature that analyses this phenomenon, aside from texts that denounce information pollution and the risks of deepfakes, or papers and diagrams that describe the various model training methodologies. The first steps of AI Image Generation were made in the 1960s. Harold Cohen in 1973 developed the Aaron AI system, which could generate black-and-white drawings. But to identify something similar to the current tools, we need to jump forward to 2014, with the publication of the first GAN by Ian Goodfellow. In 2021 OpenAI publishes Dall-E, ushering in the mass deployment phenomenon that is still ongoing.
Italiano