In a 1991 interview for the book Songwriters on Songwriting, Bob Dylan made a controversial statement: “The world don’t need any more songs,” he said. “If nobody wrote any songs from this day on, the world ain’t gonna suffer for it. Nobody cares.”
Of course, this was Dylan being typically ornery and unpredictable. After all, since 1991 he has released 13 albums and hundreds of songs, so he hardly closed up shop. And today’s Taylor Swift and Beyoncé fans waiting for the next album to drop would beg to differ about the power of brand-new music. But sometimes I wonder what Dylan would say now, with new AI music generators about to add a seemingly infinite number of songs to the music universe—including songs with voices that sound eerily similar to Dylan himself.
Emerging AI tools from Udio, ElevenLabs, and OpenAI’s Suno let people create new songs with simple prompts, while other AI products offer voice cloning to birth tunes like the viral "Heart on My Sleeve," which mimicked the vocals of Drake and the Weeknd. While details about the data that trained these AI tools are sparse, there is plenty of reason to believe that they are trained on copyrighted music. For example, the founder and CEO of Udio, David Ding, told me last week that the tool was trained on the “best quality music that's out there,” adding that “we are talking with different rights holders” about licensing data going forward.
Human songwriters are speaking up: Yesterday Sony Music Group, which owns such well-known labels as Columbia Records, RCA Records, and Epic Records, and represents artists including Adele and Beyoncé, warned hundreds of companies not to train artificial intelligence models on its content without permission. And artists such as Nicki Minaj, Billie Eilish, and Stevie Wonder pushed back on the practice last month by signing an open letter saying, “This assault on human creativity must be stopped.”
The letter hit close to home. In addition to covering AI as a Fortune reporter, I’m a part-time songwriter who has recorded and released a half-dozen albums and almost a hundred original songs over the past two decades. People often ask me how I feel about AI models training on the articles I write (which yes, I know, is ironic as an AI reporter), but to be honest I feel far more emotional about the idea of AI training on my songs—or those of songwriters I admire.
Of course, in some ways I agree with Bob Dylan: I know no one really cares that I poured my creative soul into crafting the perfect chorus, or shared my deepest feelings in a lyric that I sweated over for days, or spent hours in the recording studio perfecting vocal tracks. I’m just one of millions of people around the world who enjoy the process of songwriting as a way to express myself and share whatever talent I’m perceived to have with anyone who cares to listen.
But as many creatives—writers, artists, designers, photographers, and musicians—have experienced over the past couple of years, there is a humbling, even humiliating feeling knowing that your most personal, thoughtful work is being scraped from the web to train artificial intelligence models that can spit out songs that are getting better and better all the time. I know many working musicians (including several who signed the open letter) who are legitimately fearful of how their work is being devalued in the age of AI. These are often independent or mid-level artists who already have had to adapt to an era of laughably low streaming royalties and the reduction in live performing opportunities since the pandemic. And I can only imagine how music’s biggest stars must feel when they discover AI-generated versions of their voices, likenesses, and songwriting styles all over the web.
Of course, as an AI reporter and musician, I can also embrace the exciting creative opportunities that will come from AI-generated music. For instance, Metro Boomin is an influential record producer who signed the open letter pushing back on AI-generated music, but also accidentally sampled an AI-generated song himself when creating a humorous Drake diss track called BBL Drizzy that quickly went viral over the past week. It’s easy to see the potential for new and unique AI-enabled music experiences. But even beyond the legalities around AI model training, I want to believe—and fervently hope—that there will always be a place for humans to not only write an original song that connects with an audience but make a living from doing so.
Udio CEO Ding believes there will be. “People love their artists,” he said. “They are attached not only to the music, they also attach to the stories and personalities that are behind the music.”
Strangely, I look back to that same 1991 Bob Dylan interview for sustenance. Yes, perhaps “the world don’t need any more songs.” But Dylan made sure to add a caveat: “Unless someone’s gonna come along with a pure heart and has something to say,” he explained. “That’s a different story.”
Sharon Goldman
sharon.goldman@fortune.com
Want to send thoughts or suggestions to Data Sheet? Drop a line here.
Today's edition of Data Sheet was curated by David Meyer.
