In a groundbreaking move, Google has introduced MusicLM, an experimental AI tool designed to seamlessly transform textual descriptions into captivating musical compositions. Available through the AI Test Kitchen application on various platforms such as web, Android, and iOS, MusicLM provides users with the extraordinary ability to input prompts like "soulful jazz for a dinner party" or "craft an industrial techno sound with hypnotic undertones." Subsequently, the tool generates multiple versions of the specified song, granting users an immersive and creative experience.

Customization Beyond Imagination

The brilliance of MusicLM lies in its versatility. Users have the power to tailor their musical creations by specifying instruments, ranging from "electronic" to "classical." Moreover, the AI tool allows users to capture the desired "vibe, mood, or emotion," enabling them to refine and fine-tune their MusicLM-generated tracks. This unprecedented level of customization paves the way for endless creative possibilities.

Ethical Considerations and Collaborative Endeavors

While Google initially indicated no immediate plans for MusicLM's release, the landscape has evolved. Google has actively engaged with musicians and facilitated workshops to harness the potential of this technology in empowering the creative process. Notably, the version of MusicLM within the AI Test Kitchen is deliberately designed not to generate music featuring specific artists or vocals, showcasing Google's commitment to ethical considerations.

Navigating Copyright and Artistic Integrity

The realm of generative music is not without its challenges. A significant ethical concern revolves around copyrighted material being inadvertently incorporated into the AI-generated songs due to the training data. This dilemma was highlighted when Jay-Z's record label pursued copyright claims against a YouTube channel that employed AI to produce Jay-Z covers. Although YouTube eventually reinstated the videos, the legal ambiguity surrounding deepfaked music remains.

The landscape becomes murkier when homemade tracks, created using generative AI, gain popularity by replicating familiar sounds with striking authenticity. Music labels have taken a proactive stance by flagging such tracks to streaming platforms, citing concerns related to intellectual property. Notably, Spotify recently removed a substantial number of AI-generated songs from the startup Boomy following a complaint from Universal Music Group.

The Copyright Conundrum

A thought-provoking whitepaper authored by Eric Sunray, currently a legal intern at the Music Publishers Association, raises pertinent questions about the legality of AI music generators like MusicLM. The paper asserts that these AI systems violate music copyright by weaving coherent audio from the source material used in their training, potentially infringing upon the United States Copyright Act's reproduction right. The essence of AI technology such as MusicLM involves learning from existing musical compositions to replicate similar effects, a concept that poses challenges and discomfort for some artists.

Legal Battles and Future Clarity

The path ahead promises a clearer understanding of the intricate relationship between AI-generated music and copyright. Ongoing lawsuits in the legal arena are poised to shape the landscape of music-generating AI. One particularly significant case revolves around the rights of artists whose creations are employed to train AI systems without their awareness or consent. The outcome of these legal battles is likely to have a profound impact on the future of AI-generated music.

Google's MusicLM stands as a testament to the remarkable synergy between cutting-edge AI technology and the world of music. As this innovative tool continues to evolve, it raises essential questions about copyright, creativity, and the ethical boundaries of AI-generated artistry. While challenges persist, the collaborative efforts between technology giants, artists, and legal experts offer a glimpse of the potential harmony that could be achieved between AI and the realm of musical expression.

You can sign up to try it in AI Test Kitchen on the web, Android or iOS. Just type in a prompt like “soulful jazz for a dinner party” and MusicLM will create two versions of the song for you. You can listen to both and give a trophy to the track that you like better, which will help improve the model.

Examples