Artificial intelligence allows anyone to produce professional-sounding music in virtually any genre. Its use is surging in music and has caught the attention of major industry groups. In June, the Recording Industry Association of America, Universal Music Group, Sony Music Entertainment and Warner Music Group teamed up to sue popular AI music apps Suno and Udio, accusing them of copyright infringement.
Lawsuits like this could help safeguard the rights of musicians and record labels -- though their effect could take years. In the meantime, AI could well present more opportunities than challenges for musicians. "Most musicians I know aren't afraid of their art being replaced by AI," said Sum Patten, a creative director at the agency Glow and an adviser to the AI 2030 initiative, which promotes responsible AI practice. "It's pretty clear at this point that AI won't be able to replicate the magic that a skilled and seasoned musician can accomplish."
AI-generated songs lack the fluidity of music created by humans. But musicians who experiment with AI can give themselves an edge in an evolving industry. AI can expedite their own creative work and provide inspiration.
To understand how, consider the way Suno generates music. The user enters a text prompt, as in the example below.
Suno generated all that you hear and see here -- the lyrics, the song, the album cover -- within 20 seconds of receiving my text prompt. The song won't be a summer hit, and its cover won't win awards, but the package is pretty advanced for something created from 15 words. It showcases AI's potential to help with the creation of music, lyrics and even album art.
The app allows users to add their own lyrics and select specific styles and instruments. In the next example, I asked ChatGPT to generate the lyrics.
Suno ignored my request for female vocals, but you get an idea how we can keep customizing the song.
To a trained ear, these songs are relatively easy to recognize as AI-generated. "They're all very 'mid' in timbre, conception, performance and structure," said Eric Lyon, a composer, computer musician and professor at Virginia Tech. "The synthetic guitar solo is unnaturally regular. ... There are noticeable artifacts in the vocals."
Recognizing all these weaknesses of AI-generated music, Lyon has been using Suno mainly to inspire pieces he composes himself. As an experiment in AI-human interaction, he recently asked the app to create a parody of an atomic catastrophe. Here's what AI came up with:
"The cheerfulness of the music and text was completely incongruous for the subject of nuclear war," Lyon said. "The melody was catchy. The phrase structure was slightly asymmetrical. It suggested so many different musical ideas."
The AI song lasted 26 seconds. Lyon developed that into a six-minute composition -- beginning with a bass solo derived from the AI song to add complexity and foreshadowing.
AI is incapable of creating the magic that proficient human musicians can achieve. But when guided and curated by artists, it can help generate masterpieces. In the example below, a group of musicians and scientists trained AI on Beethoven's music, then used it to complete the great composer's unfinished 10th Symphony. The AI model generated multiple possibilities, and the musicians chose the contributions that made sense.
Working together, they expanded Beethoven's original 200 notes to a 40,000-note symphony. Audiences who listen to the result struggle to distinguish Beethoven from AI.
"When you use AI in the creative domain, it is just a tool," said Ahmed Elgammal, a computer science professor at Rutgers University who led the 10th Symphony project. Humans and AI play complementary roles. "AI has the ability to surprise us, but the human artist is the one who put that into a human context," he wrote in an email.
Technology has for many decades influenced the ways in which music is created and consumed, transforming the industry and driving it forward. "Whether it was during the rise of electronic music or the evolution of hip-hop, an art movement supported by technology, there was always a skeptical crowd that was like, 'That's not real music. Oh, no. Music is dying,'" Patten said of the AI 2030 initiative. "They were eventually silenced by a seismic wave of new artistic expression. And everyone survived."
While it's easy to imagine how AI can empower both seasoned musicians and amateur music enthusiasts, the tools are not yet powerful enough to unsettle musicians' livelihoods. "Most of the people I've worked with are pretty excited and curious where it goes," said Craig Arthur, who has shared stages with Ye and other famous artists over a 27-year career as a DJ. He has used AI tools such as Serato to blend instrumental elements from various songs, a luxury once reserved for elite music mixers. "AI has opened up possibilities for DJs," he said.
As the music industry has shown many times, open-minded artists who are quick to embrace technology stand the best chance to benefit from innovations. Rather than avoid AI, today's experimenters are engaging in critical conversations about how best to use it. This keeps them involved in AI's development, so that they can help ensure the technology compliments human creativity.
To close, let's listen to AI's own pitch on why musicians should use it.
