Illustration 264334909 © Pojoslaw | Dreamstime.com
With artificial intelligence taking the tech world by storm, protecting digital artwork has become an increasingly complex challenge. To address this, a team of researchers led by Ben Zhao from the University of Chicago has developed a new tool known as ‘Nightshade’.
Nightshade operates by subtly altering the pixels of an image, a process colloquially referred to as “poisoning” for AI training. These alterations, while completely imperceptible to the human eye, are enough to confuse algorithms.
By now, I’m guessing most have already seen the news on our new project, Nightshade. Lots of artists sharing it, but here’s the article from MIT Technology Review (thank you to the wonderful @Melissahei), and a thread explaining its goals and design. https://t.co/N01ThDT5r7
— Glaze at UChicago (@TheGlazeProject) October 24, 2023
The real impact of Nightshade comes into focus when it floods AI systems with a multitude of these “poisoned” images. Such an influx can contaminate the training data, leading to erroneous results from AI image generators. While an affected solitary image may have limited impact, thousands of such subtly altered visuals could render AI tools ineffective.
The challenge intensifies as AI image generators frequently draw fresh samples from the vast internet landscape, making it a formidable task for companies to detect and eliminate these tainted images from their training datasets.
ok ok, 1 last tweet I promise.
I realized the most surprising result was not included in the MIT TR article. You can read the details in the paper (fig17), and I will just leave the figure here. FIN/ pic.twitter.com/zeDDlHbVEO
— Glaze at UChicago (@TheGlazeProject) October 24, 2023
Previously, the team devised another tool called ‘Glaze’ that conceals an artist’s distinct style. Nightshade is set to be integrated into Glaze, opening the doors to being made open source.
This move aims to empower others in safeguarding the work of digital artists, reinforcing the role of technology as both a challenge and a solution in the ever-evolving world of art and AI.
[via Mashable and Venture Beat, cover illustration 264334909 © Pojoslaw | Dreamstime.com]
Recent Comments