Should AI sing?

Chyung Eun-ju

Joel Cho

By Chyung Eun-ju and Joel Cho

The recent news of Google and Universal Music Group engaging in discussions to negotiate a music generator utilizing AI technology has intensified the ongoing debate surrounding AI-generated music. These two corporate giants are exploring the development of a tool that would enable users to create music featuring the voices of their favorite artists. The idea is to basically grant artists the ability to "license" their voices, allowing people to purchase the rights to generate personalized music that is technically sung by the artist of their choosing.

South Korea's largest K-pop label, HYBE, the powerhouse behind the hit boy band BTS, has also delved into the realm of AI-driven music. HYBE utilized SuperTone's cutting-edge technology to release MIDNATT's song in six different languages. The AI technology took MIDNATT's voice and used recitals of the lyrics by the respective native speakers in order to meticulously adjust the artist's pronunciation to ensure it sounded correct.

Wooyoung Chung, the CEO of HYBE IM, shed light on this groundbreaking approach. They deconstructed sound into various components, including pronunciation, timbre, pitch and volume. This simultaneous release of the song in six different languages represents a global first in the music industry.

MIDNATT, the artist at the center of this AI-driven musical experiment, elaborated on how AI has expanded his artistic horizons, particularly in overcoming language barriers. While AI undoubtedly offers numerous benefits, it also sparks a vital conversation about the evolving landscape of music creation and the role of AI in this creative process.

On the other side of this debate, however, we have artists whose voices have been used against their consent. Earlier this year, an individual imitated the voices of Drake and The Weeknd and made a successful song "Heart on My Sleeve." Such AI-generated tracks often find their way onto popular platforms like TikTok, Spotify and YouTube, creating a gray area within the music industry. This phenomenon bears resemblance to previous disruptive technologies such as synthesizers, samplers and file-sharing services like Napster. Concerns regarding AI technology in music persist, particularly in terms of its potential to learn and potentially dilute copyrighted material, raising apprehensions within the music business.

Universal Music Group, the largest major label representing artists like Drake and The Weeknd, took action to remove the aforementioned song. In a statement, the company shed light on the broader implications of this technology, questioning whether stakeholders in the music industry should prioritize human creativity and artists' rights or permit the proliferation of deep fakes and fraud that deprive artists of their rightful compensation.

Artists and their labels currently maintain that the emotional connection fans share with genuine artists will set them apart from AI-generated imitations, even if AI can replicate their emotional themes and musical styles. Nonetheless, the potential impact extends beyond impersonation. AI-powered royalty-free music generators can now create rap beats, commercial jingles or film scores, posing a threat to the livelihoods of working musicians in an already precarious industry.

The generation of music through AI technology also raises concerns that echo those discussed in relation to other forms of AI-generated work, particularly in the debate surrounding artwork and the issues of authorship and ownership of rights arising from such creations.

On this matter, the U.S. District Court recently decided that AI-generated artwork is not apt to receive copyright protection. This decision was made by Judge Beryl A. Howell in the legal action filed by an individual against the U.S. Copyright Office, where the plaintiff argued that the U.S. Copyright Office unjustly denied his claim for the copyright of an artwork generated by AI. Judge Howell bases her decision basically on the reasoning that "human authorship is an essential part of a valid copyright claim."

This recent position of the U.S. Court translates to a precedent that supports the original artist's claims over the use of the likeness of their voices mimicked by AI programs.

On the other hand, this recent decision in the U.S. did not end the discussion on the necessity of adaptation on the legal dispositions to protect work and the complexities of defining authorship in this new era of AI generation. Judge Beryl A. Howell recognized in her decision that "Undoubtedly, we are approaching new frontiers in copyright as artists put AI in their toolbox to be used in the generation of new visual and other artistic works. The increased attenuation of human creativity from the actual generation of the final work will prompt challenging questions regarding how much human input is necessary to qualify the user of an AI system as an "author" of a generated work, the scope of the protection obtained over the resultant image, how to assess the originality of AI-generated works where the systems may have been trained on unknown pre-existing works, how copyright might best be used to incentivize creative works involving AI, and more."

So currently, there isn't a clear-cut law that explicitly states you can't train AI to replicate someone's voice and create a synthetic version of it. This is all a gray area in the realm of intellectual property. Interestingly, as we mentioned above, YouTube and Universal Music Group announced the Music AI Incubator, where artists and songwriters and producers use AI music tools. Elaborating more on this collaboration, an intriguing aspect is that YouTube had also hinted at making a system similar to Content ID, a tool used to identify and manage copyrighted content on the platform. What makes this development especially fascinating is its potential application to AI-generated tracks. YouTube is considering a mechanism that would quickly detect AI-generated content such as the fake Drake song as soon as it is uploaded. Then, the royalties would be directly distributed to Drake or his record label. This proposition introduces even more complex questions and challenges concerning intellectual property, copyright and compensation.

However, while YouTube is trying to protect an artist's voice, Google, which owns YouTube, is not extending similar protection to the several publishers that produce text. Google uses all the text to train AI models that create synthetic text, but the newspapers or blogs are not being compensated the way singers would be. This difference of Google and YouTube between singers and publishers is due to who has leverage in business ― for instance, YouTube needs music labels on their good side to make it easier for users to upload videos, while the media industry needs Google to keep them afloat ― perhaps a system like Content ID will be beneficial for all industries that are having their work synthesized.

As AI continues to influence the creative landscape, the intersection of technology, intellectual property and creative rights becomes increasingly complex, but the creative possibilities are becoming even more captivating.

Chyung Eun-ju (ejchyung@snu.ac.kr) is studying for a master's degree in marketing at Seoul National University. Her research focuses on digital assets and the metaverse. Joel Cho (joelywcho@gmail.com) is a practicing lawyer specializing in IP and digital law.