According to the Harvard Business Review, the introduction of generative AI image applications such as Midjourney and Stable Diffusion has disrupted the lives of artists everywhere. These tools, which are readily available to the masses, learn from the vast collections of artists' works available online. They repurpose them and create unique images based on text prompts submitted by individual users.
The apps also allow anyone to take specific digital images of original artwork and generate more versions, using the artist's hard-earned art styles and original ideas to their desire. It can be seen as copyright infringement done en masse, and yet wide public acceptance, known or unknown, has meant artists are struggling to fight back.
More than that, this has also led to artists directly losing their jobs, commissions, and basic income, as their clients now prefer to get AI-generated images at a much quicker pace and lower cost. However, a recent initiative by the Glaze Project, in collaboration with Cara, an online artist portfolio and social platform, has given artists access to a tool that allows them to take back control from AI, reflecting the United Nations Sustainable Development Goal for Decent Work And Economic Growth.
Glaze, coming out of the University of Chicago’s SAND Lab, is a research project led by Shawn Shan, a fourth-year Ph.D. student. He developed Glaze alongside his professor of computer science and renowned computer scientist, Ben Zhao. What they created is something that can be summarized as a filter for artists to apply over digital images of their works. The filter fools generative AI applications into thinking that the source images have properties different from how humans view them.
For example, an artist’s hyper-realistic painting of a dog will have detailed anatomy and fur for the human eye. Still, when Glaze is applied to a digital image of the painting, an AI application will register it as a cartoon drawing of a dog with no fur and non-specific anatomy.
This means that images generated from “glazed” artworks will look very different from their source pieces. They will not have the same line marks, shapes, shading, or even colours. Glaze ensures that the artist’s styles and, therefore, rights and livelihood are protected.
The success of the innovation is evident as Glaze earned a spot on Time magazine's list of the Best Inventions of 2023, and its creator, Shan, has been recognized by making it to the 2024 Forbes 30 Under 30 list. However, the project recognizes that there’s still a lot of work that needs to be done before the tool can be accessible to the wider public.
The work required comes in three main forms: first, the Glaze filter is still somewhat visible on artworks that have flat colours and smooth backgrounds. Second, there is a possibility that future AI algorithms may be able to override Glaze’s properties. Third, although Glaze is free to download, it requires artists to learn how to properly run the software, meaning less technologically savvy artists will feel discouraged from using it.
WebGlaze is an easy-to-use online version that was launched to help bypass the third problem, but it still requires Glaze developers to individually screen users to ensure that they are indeed human artists and not third parties that would benefit.
The Cara x Glaze initiative is also a solution, where Cara’s existing and new users of its art portfolio platform can easily apply Glaze to digital images of their works, while the Glaze Project team can breathe easy, knowing their tool is indeed being used by verified human artists.
Several artists, including photographer Jingna Zhang, known for blending Asian and Western art styles in her fine art and fashion works; photographer Bill Saltzstein, specializing in capturing animals and natural landscapes; comic artist Sarah C. Andersen, renowned for her playful and relatable comic strips; and digital artist Karla Ortiz, recognized for fantasy concept art and illustrations, have utilized Glaze.
This diverse range of applications demonstrates how Glaze can be employed across a wide spectrum of artworks and art styles, with the hope of collectively safeguarding artists from AI-generative applications.