Glaze Protection from AI
TLDRThe panel discussion, moderated by Mia Araho, addresses concerns over generative AI's impact on the art industry. Panelists include John Lamb, an advocate for artists' rights in the digital space; Ben Jiao, a computer science professor at the University of Chicago; Steven Zapata, a designer and illustrator focusing on education; and Carla Ortiz, a fine artist known for her work in the film industry. They discuss the launch of 'Glaze,' a tool developed to protect artists' work from being misused by AI models. Glaze alters images in a way that is unnoticeable to humans but disrupts AI training, preventing the AI from learning and replicating the artist's style. The conversation also touches on the potential of 'Nightshade,' a more aggressive approach to combat AI misuse by poisoning AI training data. The panelists emphasize the importance of artists using tools like Glaze to protect their work and the ongoing efforts to advocate for artists' rights in the face of advancing AI technologies.
Takeaways
- 🎨 The panel discusses the impact of generative AI on the art industry, focusing on concerns about AI's ability to replicate artists' work without consent.
- 🤖 John Lamb initiated the hashtag #createdontscrape to raise awareness and provide a safe space for artists to share their work without fear of AI misuse.
- 👨💼 Ben Jiao, a computer science professor, highlights the importance of building systems to protect AI from misuse and the ethical considerations surrounding AI technology.
- 🖌️ Steven Zapata, a designer and illustrator, expresses the need for advocacy and protection of artists' rights in the face of AI's potential to disrupt the creative industry.
- 🧙♀️ Carla Ortiz, a fine artist, shares her experience as a plaintiff in a class action suit against companies using generative AI and her efforts in advocating for artists' rights.
- 🤖 Generative AI models are trained on massive datasets scraped from the internet, including countless artists' works, raising questions about copyright and fair use.
- 🚫 The panelists argue that AI-generated art is fundamentally different from traditional tools and technologies, as it consumes and mimics the original creations of artists.
- 🛡️ 'Glaze' is a tool developed to protect artists' work from being trained on by AI systems, by altering images in a way that is unnoticeable to humans but disrupts AI training.
- 🔥 'Nightshade' is a more aggressive approach, designed to 'poison' AI training by introducing incorrect data that teaches the AI system to make mistakes.
- 🌐 Web Glaze is a user-friendly platform that allows artists to apply the Glaze protection to their artwork without needing powerful computing resources.
- ⚖️ The ongoing discussions and potential legal challenges, such as those by the FTC and class action lawsuits, could influence the future of how AI can use artists' work.
Q & A
What is the main concern of the panelists regarding generative AI?
-The panelists are primarily concerned about the impact of generative AI on artists and their work. They discuss issues related to the unauthorized use of artists' work in training AI models, the lack of compensation for artists, and the potential for AI to mimic and thereby devalue original artistic creations.
How does John Lamb describe the initial reaction to his tweet humanizing the impact of AI on artists?
-John Lamb mentions that his tweet, which aimed to humanize the suffering of artists due to generative AI, received an overwhelming response with about 60,000 retweets almost overnight. This indicated to him that people do care about the issue, they just need to understand the human side of it.
What is the role of Ben Jiao in the context of the panel?
-Ben Jiao is a professor of computer science at the University of Chicago. He has been working on systems to protect AI from misuse and has become involved in the conversation about the impact of AI on artists. His research and insights are being utilized to develop tools like Glaze, which aim to protect artists' work from being exploited by generative AI.
How does Steven Zapata's experience as an artist and educator influence his perspective on AI?
-Steven Zapata's perspective is influenced by his direct interaction with students who were stressed about the advent of AI technologies like Dolly and stable diffusion. His advocacy work and the video he created for his YouTube channel reflect his concern about the impact of AI on the art industry and the need to defend artists' rights.
Carla Ortiz describes her work in the film industry and her reaction to generative AI. What actions has she taken?
-Carla Ortiz has worked as a fine artist, concept artist, and illustrator in the film industry for over a decade. Upon realizing the potential misuse of generative AI and its impact on artists, she became one of the first plaintiffs in a class action suit against companies like stable diffusion and Deviant Art. She also advocates for artists' rights and protections alongside organizations like the Concept Art Association.
How does the Glaze technology work to protect artists' work from being used in AI models?
-Glaze is a tool developed to make small changes to images that are not perceptible to the human eye but significantly alter how AI interprets the image. When an artist's work is run through Glaze, the output image contains alterations that teach AI models something different about the artist's style, thus disrupting the training process of AI models that aim to mimic the artist's style.
What is the significance of the term 'poison attack' in the context of Nightshade?
-A 'poison attack' refers to the strategic introduction of incorrect or misleading data into a machine learning model's training set with the intention of degrading the model's performance. In the context of Nightshade, it is used to disrupt the base model training of AI systems by introducing images that appear one way to humans but are labeled and processed differently by the AI, thereby 'poisoning' the AI's understanding and output.
How does the panel view the future of generative AI in relation to jobs in the creative industry?
-The panel expresses concern that generative AI could lead to job losses in the creative industry. They argue that while AI might initially seem to increase productivity, it could ultimately replace human workers, leading to unemployment and a concentration of profit among a small group of managers and shareholders.
What is the potential impact of FTC decisions and court cases on existing AI models?
-FTC decisions and court cases could potentially require companies to destroy AI models found to rely on ill-gotten data. This could disrupt the current use of existing AI models and create uncertainty for companies relying on these technologies, as they may be forced to cease using models that are found to infringe on copyrights or other legal protections.
How does the panel suggest that artists can contribute to the fight against unauthorized use of their work by AI?
-The panel encourages artists to use tools like Glaze to protect their work. By applying Glaze to their artwork, artists can make it more difficult for AI models to learn and mimic their unique styles. The panel also emphasizes the importance of advocacy and collective action to raise awareness and push for legal protections.
What are the potential economic implications of Nightshade for companies relying on generative AI?
-Nightshade could make it more challenging, time-consuming, and expensive for companies to develop new AI models, as they risk poisoning their models with incorrect data. This could lead to a slowdown in the development of new AI technologies and potentially impact the share value of companies if they are unable to continue building and improving their models.
Outlines
🎤 Introduction to the Panel on Generative AI and Artistic Concerns
The panel, moderated by Mia Araho, discusses the impact of generative AI on artists. Concerns are raised about the safety of posting artwork online due to AI technologies like Dolly and stable diffusion. The panelists, including John Lamb, Ben Jiao, Steven Zapata, and Carla Ortiz, introduce themselves and share their motivations for advocating for artists' rights in the face of AI's challenges.
🖌️ The Concerns of Artists with Generative AI and the Creation of 'Create Don't Scrape'
John Lamb shares his personal journey in advocating for artists, starting with a tweet that humanized those suffering from AI's impact on their work. He discusses the creation of the hashtag 'create don't scrape' and the website crap.com, which compiles information on generative AI's effects on artists. Ben Jiao, a computer science professor, explains his shift in focus towards the ethical implications of AI, including its misuse and the protection of individuals from unauthorized facial recognition.
🤖 How Generative AI Works and Its Impact on Artistic Styles
Steven Zapata explains his initial research into AI and its impact on art, leading to his advocacy work. Carla Ortiz discusses her experience as a concept artist and her involvement in a class action lawsuit against companies utilizing generative AI. The panelists differentiate generative AI from other artistic tools and technologies, emphasizing its unique consumption of other artists' work.
🛡️ The Development of 'Glaze' as a Protective Tool for Artists Against AI
The panelists delve into the creation of 'Glaze,' a tool developed by Ben Jiao's team at the University of Chicago to protect artists' work from being misused by AI. They discuss the collaboration between artists and technologists in developing Glaze and the importance of maintaining trust within the artistic community.
🧩 The Technicalities of Glaze and Its Effectiveness Against AI
Ben Jiao explains the technical aspects of how Glaze works, emphasizing its ability to alter images in a way that is indiscernible to humans but disrupts AI's understanding of the artwork. He also discusses the limitations of other methods like pixel smudging and the unique approach Glaze takes to protect art from AI training models.
🌟 The Future of Artistic Protection: Nightshade and Web Glaze
The panelists discuss the upcoming tool 'Nightshade,' which complements Glaze by targeting the base model training of AI. They also talk about the development of Web Glaze, which provides online access to Glaze's protective features, and the importance of community involvement in using these tools to safeguard artists' works.
💸 Monetization and the Artistic Community's Struggle with AI
The panelists address questions about the business model behind Glaze and Nightshade, confirming that these tools are offered free of charge to artists. They discuss the challenges of funding and sustaining these initiatives, emphasizing the importance of grants and the potential for donations through a development office.
🤝 Collaboration Across Creative Industries in the Face of AI
The panelists highlight the importance of unity among different creative industries facing AI challenges. They mention an upcoming panel discussing union actions regarding AI and share their participation in an FTC listening session panel, which included representatives from various creative industries.
🌐 The Potential of Generative AI and Its Ethical Considerations
The panelists debate the potential positive applications of generative AI, with some arguing that its benefits are overshadowed by the potential harm it poses to artists and creative industries. They discuss the need for continued action and advocacy to protect artists in the face of advancing AI technologies.
⚖️ Legal and Regulatory Responses to AI's Impact on Art and Artists
The panel concludes with a discussion on the legal and regulatory landscape surrounding AI and art. They mention upcoming FTC decisions and court cases that could potentially affect the use of existing AI models and emphasize the importance of artists taking action now to protect their work and livelihoods.
Mindmap
Keywords
💡Generative AI
💡Mid Journey Dolly
💡Stable Diffusion
💡Create Don't Scrape
💡AI Litigation
💡Glaze
💡Nightshade
💡Opt-Out
💡Concept Art Association
💡AI and Employment
💡Unions and AI
Highlights
The panel discusses concerns over generative AI's impact on artists and their work, focusing on tools like Dolly and stable diffusion.
John Lamb initiated the hashtag #createdontscrape to humanize artists suffering from AI's misuse of their art.
Ben Jiao, a computer science professor, has been working on systems to protect AI from misuse and has now shifted focus to protect artists from AI infringement.
Steven Zapata, a designer and illustrator, has been advocating for artists' rights and educating about the dangers of generative AI through his YouTube channel.
Carla Ortiz, a fine artist, has become a plaintiff in a class action suit against companies using generative AI without artist consent.
Generative AI tools work by using natural language prompts to generate images that mimic the styles of artists without their permission or compensation.
The panel emphasizes the difference between generative AI and other art tools, noting that AI consumes and mimics the original work of artists.
Glaze is a tool developed to protect artists' work from being used in generative AI models by altering images in a way that is unnoticeable to humans but disrupts AI training.
Nightshade is a more aggressive tool that actively 'poisons' AI training by introducing incorrect data that teaches the AI incorrect outcomes.
The development of Glaze and Nightshade involves extensive collaboration with artists to ensure the tools are effective and user-friendly.
Glaze has been downloaded over 1.3 million times, indicating a significant number of artists are utilizing the tool to protect their work.
The team is working on a new version of Glaze that will be more effective and better at preserving the original style of the artwork while adding protection.
Web Glaze is a free service that allows artists to use the Glaze technology without needing their own powerful hardware.
The panel encourages artists to Glaze their art before sharing it publicly to protect it from being used in generative AI models.
The Concept Art Association is actively seeking funding for advocacy work to support artists in the face of generative AI threats.
The panelists discuss the importance of unionizing and collective bargaining to protect creatives from the negative impacts of AI.
Generative AI has potential positive applications, such as synthetic voice generation for bringing lost loved ones back to life, but these are limited.
The future of generative AI models is uncertain due to upcoming FTC decisions and court cases that may affect their legality and use.