Glaze Protection from AI

The panel discussion, titled 'Create Don't Scrape Glaze', focuses on the impact of generative AI technologies like MidJourney, DALL-E, and Stable Diffusion on the art community. The moderator, Mia Arahao, introduces the session, emphasizing the concerns artists have about their work being used without consent by AI technologies. The panel consists of four activists who have been vocal about protecting artists' rights and intellectual property in the face of these emerging technologies. They each share their journey into activism, sparked by personal and communal impacts observed in the art community.

Panelist Steven Zapata discusses his initial reaction to generative AI technologies and how it led him to create advocacy content on his YouTube channel. He talks about the significant anxiety and uncertainty these technologies have introduced among emerging artists. Another panelist, Carla Ortiz, shares her proactive steps in protecting artists' rights, including her involvement in a class-action lawsuit against companies like DeviantArt and Mid Journey. She emphasizes her role in educating and mobilizing the art community to fight for their rights and the authenticity of their creative works.

Steven Zapata explains the basic functionality of generative AI, detailing how users typically interact with these models via prompts to generate images. He highlights the training process of these AIs, which involves massive datasets scraped from the internet, containing billions of images and their descriptions. This training not only uses artists' works without permission but also locks in data, making retraining difficult. The conversation underscores the ethical implications and the irreversible nature of the training data used in AI models.

The discussion shifts to how generative AI fundamentally differs from other technological advancements in art, like Photoshop or digital tools, which traditionally aid artists without replacing their creativity. Carla Ortiz argues that generative AI uniquely exploits and depends on existing artworks to function. This segment of the talk delves into the misconceptions propagated by AI proponents and contrasts them with the actual implications of AI on artists' livelihoods and creative expression.

Ben Jiao, a computer science professor, introduces 'Glaze', a technology developed in collaboration with artists to protect their works from AI misuse. Glaze modifies artwork in subtle ways that prevent AI from learning and replicating the artist’s style. This part of the discussion explores the technical underpinnings and societal implications of deploying such tools. The collaboration between the academic and artistic communities highlights a proactive approach to defending creative works against technological exploitation.

The panel introduces 'Nightshade', a new tool designed to inject errors into AI models that train on unauthorized data, effectively 'poisoning' these models. This segment explains the mechanism behind Nightshade and its strategic role in the broader fight against the exploitation of artists' works by generative AI. The conversation covers the technical aspects, potential impacts, and the ongoing commitment of the project team to ensure the tool’s effectiveness and accessibility to the artist community.

The session concludes with the panel addressing questions from the audience, offering insights into further actions artists can take to protect their works and discussing the financial model behind the anti-AI tools like Glaze and Nightshade, which are provided free of charge. The panelists encourage the audience to actively use these tools to safeguard their creative output, highlighting the importance of community and collective action in the ongoing battle against AI exploitation in the arts.