Arts and Entertainment

Fake Drake and the AI Debate

The recent influx of AI-generated songs brings implications for the future of copyright in the music industry and has started a conversation on what makes a song truly human.

Reading Time: 5 minutes

Every day, a technological takeover draws closer as new gizmos and gadgets are introduced to the world. One frightening development has been the emergence of music created by artificial intelligence: with AI growing more powerful and accessible, creators online have been utilizing the technology to create entire songs with AI vocals. A prime example is the viral “Boy’s a liar, Pt. 2” cover by AI Joe Biden and Barack Obama, a TikTok-born cultural lovechild between the commander-in-chief and incumbent princess of New York drill. Besides being an Internet milestone, the song accurately encapsulates the current state of AI music: the vocals are choppy, with emphasis awkwardly placed on random syllables (Obama’s PinkPantheress is a clumsy half-falsetto) and frequent hiccups in cadence, but the voices are undoubtedly those of Presidents No. 44 and 46. Some creators have gone past meme parodies, such as the AI Drake song “Heart On My Sleeve,” which features convincing vocals from robot Drizzy and even a feature from an AI The Weeknd. The lyrics were written and recorded by TikTok creator Ghostwriter, who transferred their voice to the likeness of Mr. Aubrey Graham through an AI voice changer software. But the song creates a believable illusion for listeners—Justin Bieber bars and emotional melodrama are prime subject matter for the 6ix God, the instrumental is trendy, and the song’s AI artists have few performance glitches—causing some to mistake the track as an official song. To produce an AI vocal track like that of “Heart On My Sleeve,” any tech-savvy person can simply record themselves performing a part and run the audio through a voice changer, then place the vocals on an instrumental. In addition to voice impersonation, the rise of OpenAI’s language model ChatGPT has opened the door to automated songwriting. Give the AI a prompt such as “Write me a rap song in the style of 21 Savage explaining the importance of getting a colonoscopy” (an actual song the A&E department produced) and it will churn out verses and a chorus lyrically consistent with your chosen artist.

But when AI shifts from harmless TikTok videos to profitable works that imitate real creators, the issue of copyright laws arises. Just after the release of “Heart On My Sleeve,” Universal Music Group (UMG), which represents Drake—as well as a third of the music market—spoke out against the use of AI to emulate an artist’s style. “We have a moral and commercial responsibility to our artists to work to prevent the unauthorized use of their music and to stop platforms from ingesting content that violates the rights of artists and other creators,” UMG explained to the Financial Times. The company successfully scrubbed the track from most streaming services, though they could not stop Ghostwriter from releasing an expanded AI Drake album, Sincerely, Aubrey, on YouTube. Some artists, however, have encouraged the use of AI in music. Canadian musician Grimes addressed the controversy on Twitter, declaring she has no problem with AI using her voice, as long as she receives compensation: “I’ll split 50 [percent] royalties on any successful AI[-]generated song that uses my voice,” Grimes stated. “I like the idea of open[-]sourcing all art.” Despite the variety of responses, the lack of legal precedent surrounding AI keeps the ownership of an artist’s voice and aesthetic undecided.

Though the online traction of AI artist interpretations is newfound, early AI pattern recognition software was already being used to generate songs in the 1950s, and deepfakes—AI systems that can mimic speech configurations—emerged back in the ‘90s. Today, creators of AI works are able to utilize these mechanics, along with writing modules such as ChatGPT, to create a new form of music. A major question this poses is if AI-generated songs can legally be considered original music, stressing when, exactly, innovation becomes pure impersonation. Controversies of this nature surfaced in 2021 regarding Andy Warhol’s Prince Series (1984), which consists of a set of silk-screened Prince portraits based on photographer Lynn Goldsmith’s 1981 photo of the iconic American singer. Arguing that the series lacked “a fundamentally different and new artistic purpose and character,” Goldsmith took the case to the Supreme Court, suing the Andy Warhol Foundation for the unlicensed use of her work. The Foundation contended that Warhol’s interpretation of the photo was transformative and that he translated Goldsmith’s work into a new, original context. While the Supreme Court originally sided with the Andy Warhol Foundation, Goldsmith’s appeal prompted a second decision in her favor on May 18, 2023, when the court decided Warhol’s piece was created without a “purpose and character… sufficiently distinct from the original.”

AI creators can argue that training AI to mimic an artist’s music is comparable to human artists taking inspiration from existing works. Ghostwriter’s album reflects clear lyrical and musical consideration created through human effort, which can be seen as transformative in the same vein as Warhol’s stills. In an article by Harvard Law, law expert Louis Tompros suggests AI creators could respond to artists like Drake and The Weeknd by arguing that “the lyrics are different, the music is different. It’s a different song, and [they] don’t have rights to this.” Like Warhol, Ghostwriter and other AI creators could claim their work to be transformative, since it introduces new elements to replicated works. However, the recent Supreme Court decision complicates this distinction: Ghostwriter’s AI Drake album can be considered blatant imitation because it directly impersonates Drake’s style and voice, innate parts of his identity.

The legal regulations with copyright concerning this form of music remain unknown, and as labels and AI creators clash, our understanding of what is transformative art versus plain impersonation may never be fully defined. Nevertheless, we can confidently expect to see music industry executives scramble to protect their artists’ voices, crossing our fingers that, when it comes to music, listeners will value humans over robots. The technology for AI music is only bound to improve; the concept of an entirely AI-generated song becomes increasingly probable as ChatGPT lyrics can be combined with software producing chord progressions, melodies, and beats. Despite this, human-created music is likely to remain superior to its AI counterparts. “When music gets cleaned up too much, listeners lose opportunities to connect their imperfections with those in the music, the human traces that might otherwise reach the ear and burrow into the heart,” long-time Bruce Springsteen fan Warren Zane remarked in a New York Times article idolizing The Boss. AI music is inherently founded on the reproduction of established human music, making it only a pale imitation of the artistic genius of our species. Language models like ChatGPT are entirely referential in nature; lyrics are written “in the style of” certain artists and follow traditional song structure norms. Voice changers can copy the timbre of artists but cannot replicate the same emotion they bring to their original records. Until AI can truly create entirely original works (if ever), the staying power of human expression that makes music so special will prevail over the robot invasion.