A technology that may prevent ‘style copying’ of image-generating artificial intelligence (AI) has emerged.

TechCrunch reported on the nineteenth (local time) that the ‘Glaze’ team, which conducts an educational research project on the University of Chicago, has developed a technology that hides images from AI’s data learning.

The appliance developed by the Glaze team is a tool that permits the artist’s work posted online to look different from the unique even whether it is caught in AI data learning.

AI, which has learned an image with this technology applied, creates a picture that’s significantly different from the unique when ordered to mimic the variety of the unique writer. For instance, on the prompt ‘Draw me an image within the variety of Van Gogh’, the unsuitable variety of painting is presented.

“There may be a fundamental difference between the best way AI perceives the world and the best way we perceive the world,” said Ben Zhao, a professor of computer science on the University of Chicago. It’s focused,” he said.

Unlike other methods, it is thought that this method can interfere with data learning without harming the unique as much as possible. Nonetheless, within the case of a straightforward style comparable to sketch, some form of noise may occur since it is difficult to hide.

Recently, many artists are feeling burdened to post their work online due to generative AI. The Glaze team said they began the project for that reason. The outcomes of this study were published on an internet thesis site, and the applying was distributed freed from charge.

Professor Zhao said, “This technology may also be identified when AI’s learning function is strengthened,” but “Big Tech is not going to attempt to provide you with a separate countermeasure for this technology at a terrific cost.”

Reporter Lim Dae-jun ydj@aitimes.com