Artists around the world have found themselves under siege by artificial intelligence (AI) that studies their work and replicates their styles without giving them credit or compensation. The unethical nature of this issue has sparked collaborations between these artists and university researchers to find ways to protect their art. One artist, Paloma McClain, felt bothered by her art being used without her consent and decided to take action. She turned to free software called Glaze, developed by researchers at the University of Chicago, to outsmart AI models. This article explores the efforts of artists and researchers to defend against AI copycats and maintain the integrity of their creative work.
The Battle Against Invasive AI Models
Glaze, a software created by the University of Chicago, acts as a shield against invasive and abusive AI models. The software has the ability to tweak pixels in a way that is indiscernible to human viewers but dramatically alters the appearance of a digital piece of art to fool AI. Professor Ben Zhao, an expert in computer science and part of the Glaze team, emphasized the importance of providing technical tools to protect human creators from software imitators. The development of Glaze was carried out rapidly to address the seriousness of the issue and alleviate the pain experienced by artists.
Generative AI giants often use data for training purposes, but the majority of digital images, audio, and text used to train AI systems are collected from the internet without explicit consent. This raises serious ethical concerns regarding the unauthorized use of creators’ work. In March 2023, Glaze was released, and it has since been downloaded over 1.6 million times, demonstrating the demand for protective measures against AI copycats. Researchers working on Glaze are now focusing on Nightshade, an enhancement that confuses AI by manipulating images. The poisoned images created by Nightshade could have a significant impact on preventing AI from imitating artists’ work.
Startup Spawning has developed Kudurru software, which aims to detect and block attempts to harvest large numbers of images from online platforms. Artists can use Kudurru to deny access or provide misleading images that taint the data being harvested, thus hindering AI from accurately learning from it. The Kudurru network has already integrated more than a thousand websites, promoting the protection of artists’ intellectual property. Spawning has also launched haveibeentrained.com, a website that enables artists to check whether their works have been fed into an AI model and opt out of such use in the future.
AntiFake: Preserving Authentic Voices
Researchers at Washington University in Missouri have developed AntiFake software, focusing on preventing AI from copying voices. The software enriches digital recordings of people speaking by adding inaudible noises that make it impossible to synthesize a human voice. This breakthrough aims to counter the creation of “deepfakes,” which are fabricated soundtracks or videos that deceive viewers by making them believe someone said or did something they didn’t. AntiFake has garnered attention from various industries, including a popular podcast that sought help to safeguard its productions.
Jordan Meyer, co-founder of Spawning, believes that the ideal solution would involve a world where all data used for AI is subject to explicit consent and appropriate payment. This vision strives to ensure that artists receive recognition and compensation for their work while preventing unauthorized use by AI systems. However, achieving this level of consent and payment may require significant advancements in data privacy and intellectual property rights frameworks.
The rise of AI copycats has created a challenging environment for artists worldwide. Recognizing the importance of defending artists’ rights, collaborations between artists and researchers have resulted in the creation of powerful software tools such as Glaze and Kudurru. These tools act as shields and obstacles, confusing AI systems and preventing them from imitating artists’ work without consent. Researchers at Washington University are also developing AntiFake software to preserve authentic voices and prevent the creation of deepfakes. While progress is being made, the ultimate goal is to establish a world where consent and payment are essential components of AI data usage, ensuring fair treatment for artists and respect for their creations.
Leave a Reply