Introduction
AI’s Disruptive Potential in Music
Björn Ulvaeus expresses both excitement and apprehension about the rapid advancements in AI, particularly its application in music. He recounts a demonstration of an AI model that “blown [him] away” with its potential to act as a “fantastic co-writer.” However, he raises critical concerns about the ethical implications of AI models being trained on existing copyrighted music. He fears it will be “very difficult to prove” if an AI-generated melody is derived from original work, especially if the vocals are different, making remuneration for original creators a huge challenge. He stresses the urgent need to “fight for the writers” to ensure they are compensated, suggesting solutions like “blanket licenses” for AI models to use catalogs for a fee, or a “user-centric” subscription model where artists are paid only when their music is actively played.
He also acknowledges the immediate challenge of AI voice replication, citing examples of AI-generated Paul McCartney vocals. While some see this as “fair use” or “original creation,” Björn questions the legal ownership of an AI-trained voice, comparing it to an imitation that should be subject to copyright or trademark. He believes there are positive uses for AI, such as enhancing old, poor-quality demos of deceased artists.
ABBA’s Songwriting and Production Process
Björn reflects on ABBA’s unique songwriting and production methods, which were remarkably analog compared to today’s technology. He describes a meticulous process in the studio where three people manually operated the mixing desk, sometimes taking ten or more passes to achieve a perfect mix, as exemplified by “Mamma Mia.”
He also details his collaborative songwriting with Benny Andersson:
- Idea Generation: They never wrote ideas down or recorded demos, relying solely on what “stuck” in their minds, believing only the truly strong ideas would be remembered.
- Structure and Arrangement: They focused on ensuring every part of a song—intro, verse, bridge, chorus—was perfect before recording.
- Lyric Writing: Björn emphasized that lyrics were written after the basic backing track was complete, as the “sound of the track” would “conjure up images” and suggest the mood for the lyrics. He cited “Knowing Me, Knowing You” as an example where the music created a clear visual narrative in his mind.
- Vocal Delivery: He explained how the ladies, Agnetha and Frida, would receive lyrics in the morning, rehearse them, and then record, maintaining the same process even for their recent reunion album.
The Enduring Sound and Legacy of ABBA
Björn admits he doesn’t know why ABBA’s music has lasted for so long, attracting young fans decades later, but he feels “humbled” by its continued popularity. He attributes ABBA’s distinctive sound to the unique blend of Agnetha’s soprano and Frida’s mezzo-soprano voices, particularly when Frida pushed her upper range. When deciding who sang lead, it was often a fair rotation, with each singer taking turns.
Regarding the ownership of AI-trained voices, Björn highlights that while labels own sound recordings, artists own their voices, raising complex legal questions about “name and likeness” in the age of AI. He acknowledges the rise of AI tribute bands but notes that ABBA’s own “ABBAtars” required years of development, an intricate story, a relevant music catalog, and the authorization of all four living members to be authentic and impactful.
Advocating for Songwriters and Metadata
Björn, as president of CISAC (the umbrella organization for CMOs like PRS, BMI, ASCAP) and co-founder of the platform “Session,” expresses a strong interest in ensuring accurate and timely payments for songwriters. He explains that current systems often lead to “black box” money because metadata about songwriters isn’t included when recordings are uploaded to streaming platforms. “Session” aims to capture “rich metadata”—including unique identifiers for co-writers and works—at the point of creation, ensuring that everyone involved in a song is correctly credited and paid. He advocates for retrofitting existing music with this metadata and believes that platforms should provide detailed information about songwriters, producers, and musicians to enrich the listener’s experience. He also emphasizes that the sheer volume of AI-generated music being uploaded requires filters to protect royalties for “real artists.”