The AI ​​lab is waging a guerrilla war against exploitative AI

Still, it’s “simplistic to think that if you have a real security problem and you try to design a protection tool, the answer should be that it either works perfectly or it doesn’t implement it,” Zhao says, citing spam. filters and firewalls as examples. Defense is a constant game of cat and mouse. And he believes most artists are smart enough to understand the risk.

Offering hope

The battle between makers and AI companies is fierce. The current paradigm in AI is to build bigger and bigger models, and there is no way around the fact, at least for now, that they require huge datasets pulled from the internet to train on. Tech companies claim that everything on the public Internet is fair game, and that it is “impossible” to build advanced AI tools without copyrighted material; Many artists claim that technology companies have stolen their intellectual property and violate copyright law, and that they need ways to keep their individual works out of the models – or at least get proper recognition and compensation for their use.

So far, the creatives aren’t exactly winning. A number of companies have already been replaced designers, copywriters and illustrators with AI systems. In a high-profile case Marvel Studios used AI-generated imagery instead of man-made art in the title sequence of the 2023 TV series Secret invasion. In another, a radio station fired its human presenters and replaced them with AI. The technology has become a major point of contention between unions and film, TV and creative studios, recently leading to a strike by video game artists. There are numerous lawsuits by artists, writers, publishers and record labels against AI companies. It will likely take years before a clear legal solution is reached. But even a court ruling won’t necessarily resolve the difficult ethical questions created by generative AI. It is also unlikely that future government regulation will do this, if it ever comes.

That’s why Zhao and Zheng see Glaze and Nightshade as necessary interventions – tools to defend original work, attack those who want to help themselves to it, and at least buy artists for a while. Having a perfect solution isn’t really the point. The researchers have to provide something Because the AI ​​sector is now moving at breakneck speed, Zheng says, this means companies are ignoring the very real harm to people. “This is probably the first time in our entire technology career that we’ve seen so much conflict,” she adds.

On a much larger scale, she and Zhao tell me they hope Glaze and Nightshade will eventually have the power to change the way AI companies use art and how their products produce it. It is shockingly expensive to train AI models, and it is extremely laborious for engineers to find and remove poisoned samples in a dataset of billions of images. Theoretically, if there are enough Nightshaded images on the Internet and tech companies see their models broken as a result, it could push developers to the bargaining table to negotiate licensing and fair compensation.

Of course, that’s still a big ‘if’. MIT Technology Review has reached out to several AI companies, such as Midjourney and Stability AI, which did not respond to requests for comment. An OpenAI spokesperson, meanwhile, did not confirm details about encountering data poisoning, but said the company takes the security of its products seriously and is continually improving its security measures: “We are always working on how we can make our systems more robust against this type of abuse. ”

In the meantime, the SAND Lab continues to explore funding from foundations and nonprofits to keep the project going. They also say there has been interest from major companies looking to protect their intellectual property (although they decline to say which), and Zhao and Zheng are exploring how the tools could be applied to other industries, such as gaming, videos or music. . In the meantime, they plan to continue updating Glaze and Nightshade to be as robust as possible, working closely with students in the Chicago lab, where Toorenent’s work hangs on another wall. Belladonna. The painting has a heart-shaped note taped to the bottom right corner: “Thank you! You have given us artists hope.”