The question of whether artists whose work contributes to training generative AI systems like ChatGPT should be compensated is a topic of debate. During a recent interview at SXSW, Peter Deng, VP of consumer product at OpenAI, hesitated to provide a direct answer. When asked about this issue, he responded, “That’s a great question.” Some audience members voiced their support for compensating artists, while Deng acknowledged their input.
OpenAI’s position is complex due to legal considerations. Generative AI systems, such as DALL-E 3 (which is integrated into ChatGPT), are trained using a vast array of publicly available data, including artwork, illustrations, and photos. OpenAI and other vendors argue that the legal doctrine of fair use protects their practice of utilizing public data for training without compensating or crediting individual artists, as long as the use is transformative.
In summary, the compensation of artists in the context of generative AI remains an ongoing discussion, with legal and ethical implications to consider.
Recently, OpenAI contended that creating useful AI models without incorporating copyrighted material would be impractical. According to the company, training AI models using publicly available internet content falls under fair use, a principle supported by longstanding and widely accepted legal precedents. OpenAI views this approach as equitable for creators, essential for innovators, and crucial for U.S. competitiveness.
However, creators hold a different perspective.
A class action lawsuit, initiated by artists including Grzegorz Rutkowski (known for work on Dungeons & Dragons and Magic: The Gathering), is currently progressing through the courts. The plaintiffs argue that tools like DALL-E 3 and Midjourney replicate artists’ styles without explicit permission, enabling users to generate new works resembling the originals, all without compensating the artists.
OpenAI has established licensing agreements with certain content providers, such as Shutterstock, and allows webmasters to prevent its web crawler from scraping their sites for training data. Similarly, OpenAI, along with some competitors, offers artists the option to “opt out” and remove their work from the datasets used to train image-generating models. Some artists find the opt-out process cumbersome due to the need to submit individual copies of each image for removal, along with descriptions.
Peter Deng, expressing his views, believes that artists should play a more active role in shaping the creation and utilization of generative AI tools like DALL-E. However, the exact form of that involvement remains uncertain.
“[A]rtists need to be a part of [the] ecosystem as much as possible,” Deng stated. He also pondered how accelerating the flywheel of art creation might impact the industry, considering that every artist draws inspiration from those who came before them.