Now artists are preventing again. Shawn Shan, a 26-year-old PhD candidate in computer science at the University of Chicago and recipient of MIT Technology Review’s 2024 Innovator of the Year award, has been instrumental in designing some of the most powerful instruments.
As an undergraduate, Shan gained experience in AI safety and privacy by participating in a project that developed Fawkes, a software designed to protect individuals from facial recognition technology. Amidst the explosive growth of generative AI, a chance encounter with distressed artists catalyzed his involvement in one of the most pivotal battles within the industry. Following their research into the impact on artists, Shan and his advisors, who had previously compiled the Innovators Beneath 35 list in 2006, as well as Heather Zheng, who featured on the 2005 list, decided to develop an application to provide assistance. Researchers collected input from over a thousand artists to understand their desires and explore how they might utilize existing protective measures.
CLARISSA BONET
Shan developed an innovative algorithm for Glaze, a cutting-edge software enabling artists to safeguard their creative processes from AI-generated mimicry. Glaze emerged in early 2023, with Shan and his team introducing another innovative tool in October – a software that applies an imperceptible “poison” layer to photographs, thereby rendering them unusable for image-generating AI models attempting to incorporate these images into their data sets? If significant amounts of poisoning are injected directly into a machine learning model’s training data, it can irreparably damage the patterns and render their outputs highly unpredictable. Algorithms function by introducing subtle alterations to photographic pixels, deliberately manipulating how machine-learning models process and understand images.
Shan describes the responses as being both overwhelming and anxious in tone. The company faced criticism from proponents of generative AI on social media, prompting several attempts to circumvent the safeguards in place.
However artists liked it. To date, Glaze has accumulated a staggering 3.5 million downloads, with Nightshade boasting an impressive 700,000 instances. Built into the popular new art platform Cara, this feature enables artists to seamlessly integrate its security measures into their work when uploading images. At the prestigious Usenix Safety Symposium, a leading conference for computer security, Glaze received two esteemed honors: the Distinguished Paper Award and the coveted Web Protection Prize.
According to Karla Ortiz, an artist who collaborated with Shan’s team on Glaze and is also involved in a class action lawsuit against generative AI companies for copyright infringement, Shan’s work has revitalized the online artistic community, allowing creators to express themselves freely once again.
Meet the remainder of this year?
.
“They’re driven by their passion for a community that’s been historically taken advantage of and exploited; they’re genuinely invested in it,” says Ortiz.
Zhao credits Shan with grasping the types of safeguards artists sought, observing how their collaborative efforts on Fawkes could inform the development of Glaze. Zhang describes Shan’s exceptional technical prowess, likening it to the strongest he’s ever witnessed; yet, what truly distinguishes him, he claims, is the uncanny ability to connect seemingly disparate dots across disciplinary boundaries. “These are the kinds of complex problems you can’t simply train for,” Zhao explains.
