Artists can now use Nightshade to fight theft, a tool that corrupts image-generating AI models.
Advertisement: Click here to learn how to Generate Art From Text
Computer scientists at the University of Chicago have developed a new tool called Nightshade that is meant to “poison” digital artwork, making it detrimental for training image-generating AI models that engage in intellectual property theft, such as DALL-E, Midjourney, and Stable Diffusion.
Where did Nightshade originate?Nightshade, developed by computer scientists from the University of Chicago under the direction of professor Ben Zhao, is part of The Glaze Project. The group had previously developed Glaze – a tool to confuse AI training algorithms by changing the way they perceive the style of digital art.
How does it Work?Pytorch, an open-source machine-learning framework, is used to tag images down to the pixel. The tags aren’t obvious to humans looking at the image, but AI models see them differently, adversely affecting how the images are used for training.
What’s the difference between Glaze and Nightshade? Glaze convinces training models that they’re seeing a different artistic style than a human looking at the image would see. For example, Glaze can convince AI models that a “glazed” charcoal drawing is actually an oil painting, while any human looking at the image would still see a charcoal drawing. Nightshade, on the other hand, convinces models that the content of the image is different from what a person would see. So, a “shaded” image could convince an AI model that a photo of a cat is actually a photo of a dog. The model would use that data to train itself, and when the user entered a text request asking for a photograph of a pet, it would receive a photograph of a dog.
What are the downsides?The glazing and shading process adds some noise to digital pictures. The amount of distortion varies from image to image, and can be altered by the user.
What is the future of image-generating programs? Not at all; and it’s worth pointing out that Glaze Project is not anti-AI. As mentioned earlier, Glaze uses open-source AI software for the image tagging. The programs were developed to create a system where users of image-generating software would need to have the permission of rightsholders in order to access unaltered training images.
What can the Glaze Project gain from this?Zhao believes that it’s not about money. On the group’s WebsiteHe writes:
Our primary goal is to discover and learn through our research new things, and make a positive contribution to the world. Ben (me) speaks for the team (but I believe that I also speak for myself) when I state that we aren’t interested in profit. There is no business plan, no subscription, hidden fees or startup.