Artificial intelligence (AI) is one of the latest innovations with technology that has become increasingly available to the public. However, this latest step forward in tech is surrounded by a lot of controversy (including whether AI will take over the world), especially generative AI models. As the name implies, this type of AI is used to generate content, such as text, images, music, etc., and are often trained using data found on the Internet. This is where many people- especially artists- have an issue with generative AI, especially image and visual art generating AI models.
What Issues Does Generative AI Pose?
Oftentimes, generative AI models “have been trained on large datasets of scraped images from online, many of which are copyrighted, private, or sensitive in subject matter” (Zhao). This training is usually done without “…[the] knowledge, consent, credit or compensation” of the artist whose work is being used (Zhao et al.). Further, “to make it worse, many of these models are now used to copy individual artists, through a process called style mimicry”, which essentially allows a user to generate a picture that looks like a specific artist’s style if that artist’s name is input as a prompt to the generative AI (Zhao et al.).
Why is this an issue? Zhao et al. explains that
Style mimicry produces a number of harmful outcomes that may not be obvious at first glance. For artists whose styles are intentionally copied, not only do they see loss in commissions and basic income, but low quality synthetic copies scattered online dilute their brand and reputation. Most importantly, artists associate their styles with their very identity. Seeing the artistic style they worked years to develop taken to create content without their consent or compensation is akin to identity theft. Finally, style mimicry and its impacts on successful artists have demoralized and disincentivized young aspiring artists. We have heard administrators at art schools and art teachers talking about plummeting student enrollment, and panicked parents concerned for the future of their aspiring artist children.
Zhao et al.
Hence, a team from University of Chicago- professors Ben Zhao, Rana Hanocka, and Heather Zheng, alumni Emily Wenger and Jenna Cryan, and Ph.D. student Shawn Shan- have created Glaze (Zhao et al., Shan).
What is Glaze?
Glaze is “a system designed to protect human artists by disrupting style mimicry” (Zhao). Glaze utilizes machine learning algorithms to compute “a set of minimal changes to artworks” which appear, essentially, unchanged to the human eye (Zhao et. al). Yet, to a machine, the changes appear drastic: “human eyes might find a glazed charcoal portrait with a realism style to be unchanged, but an AI model might see the glazed version as a modern abstract style” (Zhao et. al). This effect disrupts the generative AI’s ability to copy a given artist, since the image it sees is different from the content of the image seen by us humans. Us humans see these images visually; AI models, however, do not. They do not “look” at the images as we do, but rather they see the images as data. They see the various qualities, aspects, patterns, etc. that make up the image, but do not actually “see” it in a visual sense. For example, imagine closing your eyes, and having someone describe a painting to you; while is example is not quite perfect, it is a good exercise to help begin to understand how an AI views images.
Zhao et al. go on to explain the various questions that would naturally be proposed upon hearing this initial explanation of how Glaze works: “Why can’t someone just get rid of Glaze’s effects by 1) taking a screenshot/photo of the art, 2) cropping the art, 3) filtering for noise/artifacts, 4) reformat/resize/resample the image, 5) compress, 6) smooth out the pixels, 7) add noise to break the pattern?” One would expect that any of these seven techniques would render Glaze obsolete; however, this is not the case. Glaze “is not a watermark or hidden message (steganography)”, but rather a specific transformation of the image and its visual qualities that are near imperceivable to human beings (Zhao et. al). Thus, when an AI reads in the data for these images, it also reads in these transformations, which essentially serve to train the AI incorrectly.
For example, say you were training an AI to generate artwork to look like Pablo Picasso’s artwork.
The images you were using to train the AI, however, were Glazed. They appear, to the machine, like they were made by Salvador Dali.
As a result, the AI would produce images that resembled Dali’s style (shown above) but still claim they were in Picasso’s style.
Inspiration Behind Glaze
Zhao et al. identify three reasons that Glaze was created; two of which are that they had “been doing research in adversarial machine learning, and this was an exceptional opportunity to make a strong positive impact” (Zhao et al.). The most impactful, however, is that they believe it is important to help protect “the large majority of artists [who] are independent creative people who choose art because it is their passion, and generally barely make a living doing so”, especially since, by the time legislation is passed to help protect artists, it “might be too late to prevent generative AI from destroying the human artist community” (Zhao et al.). Overall, the primary goals of the team behind Glaze are to “discover and learn new things through our research, and to make a positive impact on the world through them” (Zhao et al.).
Where to Find Glaze
Glaze can be downloaded here: https://glaze.cs.uchicago.edu/downloads.html
Alternatively, as of August 2023, the team behind Glaze has “deployed WebGlaze, a free web service that artists can run on their phone, tablet, or any device with a browser to have their art be glazed on GPU servers we pay for in the Amazon AWS cloud” (Zhao et al.).
Both Glaze and WebGlaze are “paid for by research grants to ensure it is free for artists”. (Zhao et al.).
Read the publication for Glaze here:
https://www.shawnshan.com/files/publication/glaze.pdf
Follow Glaze on social media:
Resources & Further Reading
Shan, Shawn. Shawn Shan, http://www.shawnshan.com/. Accessed 19 Oct. 2023.
Zhao, Ben, et al. “What Is Glaze?” Glaze – What Is Glaze?, University of Chicago, 2023, glaze.cs.uchicago.edu/what-is-glaze.html.
The H.A.C.K.E.R. Project 
Leave a comment