Ed Newton-Rex, the head of audio at Stability AI, has announced his departure from the company due to concerns about the public’s response to the use of copyrighted material in training its conceptual AI solution. This decision has sparked a broader conversation about the ethics of utilizing content without permission in AI models.
Newton-Rex took to Twitter to explain his resignation, stating that he disagrees with Stability AI’s stance on training AI models with copyrighted works under the guise of “fair use.” This move garnered both support and criticism from the community.
Reid Southern commended Newton-Rex for his decision, expressing gratitude for his principled stand. On the other hand, some, like Twitter user John Harvey, questioned whether being inspired by existing content constitutes a violation of rights.
The issue came to light when Stability AI responded to a request for comment from the U.S. Copyright Office regarding conceptual AI. In a detailed 22-page statement, the company defended the use of existing content in training AI models as a morally justifiable practice under fair use and in alignment with the objectives of copyright law.
While Newton-Rex initially focused on Stability AI in his public statements, he later clarified in an interview with Decrypt that the issue extends beyond a single company. He highlighted the widespread adoption of similar strategies by numerous generative AI firms across industries, emphasizing his objection to this collective approach.
The debate revolves around the balance between authors’ rights to control the use of their work and the provisions of fair use that permit limited utilization of copyrighted material for purposes such as research, critique, or education. This delicate equilibrium seeks to safeguard both creators’ interests and the public’s access to artistic content.
Notably, authors associated with ChatGPT at OpenAI raised concerns about their work being incorporated into popular AI models without consent, leading to legal action against companies like Stability AI. However, a federal judge dismissed these claims due to insufficient evidence of copyright infringement, effectively resolving the dispute.
Newton-Rex disclosed that Stability AI’s music model was trained using tracks from the AudioSparx digital music library platform. He highlighted the mutually beneficial relationship between AI developers and content providers, emphasizing the potential for collaboration rather than conflict between generative AI and creative industries.
Despite departing from Stability AI without a concrete plan, Newton-Rex stressed the importance of ongoing discussions surrounding copyright issues and fair use in AI development. He underscored the need for transparency and ethical considerations in utilizing publicly available datasets and copyrighted material for training AI models to foster a more harmonious relationship between artists and AI designers.
