After his initial surprise, Greg Rutkowski thinks AI-generated art is a good way to reach new audiences, but it still makes him unhappy.Rutkowski is a digital artist from Poland who specializes in creating dreamy futuristic landscapes using classic painting styles.
He has previously created game illustrations for well-known games such as Sony’s Horizon: West End, Ubisoft’s Anno 1800, Dungeons and Dragons, and Magic: The Gathering. But the veteran illustrator was raided by a new world of AI-generated illustrations.
He tried searching the Internet for his own name to see if an illustration he had created had been published. But in the online results, although there are many links to his name, many were not created by him.
Rutkowski said: “It only happened in a month. So what if it was a year? Because the Internet is full of AI-created art, I may not be able to find my own. work”.
Stability.AI trained the model on the LAION-5B dataset (compiled by the German nonprofit LAION) and created Stable Diffusion.
Technologist and author Andy Baio downloads and analyzes some of the data from Stable Diffusion, LAION puts the dataset together and filters out watermarked images and unsightly images such as logo images ) to narrow the range.
Baio analyzed 12 million of the 600 million images used to train the model and found that a significant portion of them came from third-party sites like Pinterest and art shopping sites like Fine Art America.
Much of Rutkowski’s work is scraped from ArtStation, a site where many artists upload online portfolios. There are several reasons why Rutkowski’s work was picked up by AI.
Foremost among them, his dreamy ethereal style looks cool. And he also creates a lot of illustration work, and many of his illustrations are available online in high enough quality, so there are plenty of examples to choose from. An early text-to-image generator called Disco Diffusion offers Rutkowski as an example hint.
Rutkowski also added English alt text when uploading his work online. These image descriptions are useful for visually impaired people using screen reader software, and they also help search engines rank images. This also makes them easy to grab, and the AI model knows which images are relevant to the cue.