OpenAI, the company behind the viral chatbot ChatGPT, is reportedly developing a new tool capable of generating music from text and audio prompts. This news comes from The Information, which states that OpenAI is collaborating with The Juilliard School in New York City on this project. The collaboration involves Juilliard students meticulously annotating musical scores, contributing to a massive dataset designed to train the AI system. While a Juilliard representative denied these reports to Euronews Next, OpenAI has yet to publicly confirm or deny them.
This wouldn’t mark OpenAI’s first foray into music generation. Back in 2019, before ChatGPT took the world by storm, OpenAI introduced MuseNet, a neural network that could compose four-minute musical pieces using ten different instruments. A sample demonstrated MuseNet’s ability to reimagine Mozart’s iconic Rondo Alla Turca in the distinctive style of Frédéric Chopin.
Building on this foundation, OpenAI launched Jukebox in 2020. This more advanced “neural net” ventured beyond instrumental music generation, producing rudimentary vocals alongside compositions across various genres. The data used to train Jukebox encompassed not only musical styles but also artist and lyrical information, aiming for a richer output.
The potential arrival of this new AI music tool coincides with growing interest in music AI across the tech industry. Spotify recently announced a partnership with major record labels Sony Music, Universal Music Group, and Warner Music Group to develop “responsible AI products.” Spotify already utilizes AI in several features, including personalized playlists like “Daylist” and the “AI DJ,” which analyzes listening history to suggest relevant tracks.
Further fueling this trend are startups like Suno and ElevenLabs, both offering platforms centered around AI-generated music.
However, the integration of AI into music creation is facing growing scrutiny over copyright concerns. European music industry organizations, including ECSA (European Composer and Songwriter Alliance) and GESAC (European Grouping of Societies of Authors and Composers), have raised alarm bells about the EU’s AI Act. They argue that the legislation, set to take effect in 2024, fails to adequately protect creators whose work is utilized in training generative AI models without clear opt-out mechanisms or compensation provisions.
This lack of safeguards leaves music creators vulnerable to potential misuse, as evidenced by a lawsuit filed against Suno by the Recording Industry Association of America (RIAA). The RIAA alleges that Suno illegally scraped copyrighted songs from YouTube to train its AI model. Similarly, Universal Music Group has launched a copyright lawsuit against AI company Anthropic.
This rapid advancement in music generation technology underscores the pressing need for clear legal frameworks and ethical guidelines to ensure fair compensation for artists and prevent unauthorized use of their creative output.
