TLDR: Bob L. T. Sturm, an engineer and artist, reflects on creating two music albums using AI music generation platforms like Suno and Udio. His first album, “Music from the Spam Folder,” explores AI’s polished outputs from junk mail prompts, raising questions of authorship and a feeling of fraudulence. His second, “A Difficult Christmas,” intentionally features unpolished, amateur music to challenge AI’s limitations. Through this process, Sturm’s musical identity expanded to include visual elements, and he developed a new appreciation for text in music, ultimately advocating for the inclusion of “unpracticed” sounds in AI training data.
In the rapidly evolving landscape of artificial intelligence, questions of creativity, authorship, and artistic identity are becoming increasingly complex. A recent research paper by Bob L. T. Sturm delves into these very themes, offering a candid reflection on his experience creating two music albums using state-of-the-art prompt-based AI music generation platforms like Suno and Udio. This paper, titled “I made this (sort of)”: Negotiating authorship, confronting fraudulence, and exploring new musical spaces with prompt-based AI music generation, provides a unique first-person account of an artist grappling with the implications of AI in their creative process. You can read the full paper here: Research Paper.
Sturm’s journey began with his first album, “Music from the Spam Folder,” where he playfully experimented by transforming junk mail into musical prompts for AI. The surprising quality of the results led him to explore spam as a unique text source for musicalization. This initial venture raised profound questions about his role as an author when machines were generating such polished outputs.
His second album, “A Difficult Christmas,” emerged as a direct counterpoint. While the first album celebrated the AI’s ability to produce “practiced, polished, and produced” music, the second intentionally sought to create something raw, unpolished, and even amateurish – qualities that current AI music platforms struggle to replicate. This album features Sturm himself playing solo violin without mastery, embracing imperfection as a form of authenticity, and challenging the AI’s inherent bias towards perfection, likely due to its training on commercially released, professionally recorded music.
The Shifting Sands of Authorship
One of the central questions Sturm grapples with is the extent of his authorship. He likens his role in “Music from the Spam Folder” to that of someone making a mixtape, curating existing tracks rather than composing them from scratch. He identifies four main avenues for interacting with these AI platforms:
-
Prompting: Crafting text descriptions to guide the AI’s generation. While this requires skill, Sturm questions if it truly confers authorship, comparing it to ordering food at a restaurant rather than being the chef.
-
Curating: Selecting and arranging the AI-generated outputs. This involves editorial judgment, but again, he feels it’s more akin to being an editor than an author of the original works.
-
Covering/Composing: Performing or arranging AI-generated material with human input. This is where Sturm feels a stronger claim to authorship can be made, as it re-establishes a “reality” to the music beyond the AI’s simulation.
-
(Re)mixing/Sampling: Processing, combining, or transforming AI-generated audio. Depending on the extent of transformation, this can introduce significant novelty and warrant a claim of authorship.
For “Music from the Spam Folder,” Sturm’s authorship is multifaceted. He conceived the concept, chose the spam, crafted prompts, curated outputs, and performed extensive remixing and sampling on some tracks. Yet, he acknowledges that much of the album’s creative compositional choices, like vocal delivery or unexpected narrative elements, came directly from Suno and Udio, leading to a feeling of “fraudulence” or a lack of legitimation as a composer.
Confronting Fraudulence and Evolving Identity
Sturm openly discusses the discomfort of creating music that seems “beyond” his own skill or vision, generated by AI. He navigates this tension with humor and absurdity, viewing his musical endeavors as an “engineer masquerading as an artist.” This perspective allows him to be immune to “creative displacement anxiety” and to pose serious questions through his work.
The “finished product” nature of AI-generated music forced him to consider what value he, as an artist, could add. This led to an expansion of his musical identity to include visuals. He began creating synchronized videos for his AI-generated tracks, sampling various audio and visual sources, and even incorporating non-musical sounds like evangelical preachers or public council meetings. This integration of visuals became a critical part of his contribution, increasing his satisfaction and comfort in claiming authorship.
Also Read:
- Harmonizing Healing: Music-Driven Art Recommendations for Therapy
- Building AI for Everyone: A Call for Neuro-Inclusive Technology
Discovering New Musical Spaces
Sturm’s experimental approach with AI music platforms has led to unexpected discoveries. Initially, he aimed to understand the AI’s training data by prompting it with existing song lyrics, questioning if it was trained on copyrighted material. This led to a newfound appreciation for lyrics and their setting to music, a domain he had previously paid little attention to. His work with Ernst Toch’s “Geographical Fugue” and his “’round ’n ’round (traditional)” piece, which iteratively generated settings for a tragic poem, showcased the AI’s remarkable and often surprising text-setting capacities.
He also explored live performance, integrating AI-generated sound material with live video mixing and co-performers. This included a performance in a Vienna café where AI-generated music was combined with real-world actions like making a salad and taking audience orders, blurring the lines between art, performance, and everyday life.
Ultimately, Sturm’s journey has opened new musical spaces for him, fostering an appreciation for text in music, developing sensibilities about sampling and visual accompaniment, and conceptualizing his work as a form of “pseudoplunderphonics” – a plundering of music that has already been processed by AI trained on vast datasets. While acknowledging his own limitations, he sees his albums as a call to activism, challenging the AI’s bias towards polished music and inviting machines to learn from the “unpracticed, unpolished, and unproduced” sounds that represent a significant part of the world’s music.
