The Current #4: Generative AI Music

The Current is a bi-weekly series from NEA on the developments impacting consumer technology. Each installment examines a trend, disruption, or opportunity with consumer data. Posts are concise, informative, and always current.

***

The first computer-composed song predates the floppy disk, touch-tone telephone, and handheld calculator. In 1957, a mathematician and a musician at the University of Illinois composed Illiac Suite for String Quartet, the result of computer-generated integers representing notes filtered through predefined compositional rules. While a breakthrough in musical technology, its sheer inaccessibility deferred any larger questions around artistry or creation or authenticity. To start, the only known computer capable of production weighed five tons and cost a cool $5 million.

Today, consumers with no artistic training can generate studio-grade music in seconds using only natural language inputs. Streaming platforms like Spotify, Apple Music, and Soundcloud have already broadened discovery, distribution, curation – the story of the past two decades. But that disruption, from the perspective of the user, stopped short at production. No longer. Platforms like Suno, Udio, and Sonauto productize what was unimaginable just a few years ago.

LISTEN HERE. Created on 4.24.24

Considering the Impact Of AI Generated Music on the Industry and Fans

How users engage generative AI music informs a much larger story – how individuals see artistry, personalization, utility, creativity in the next chapter of consumer technology. A panel of consumers ages 18-30, living in the United States and at least familiar with the technology gave us some direction when we probed the most fundamental questions around artificial intelligence generated music.

I. Consumption

 These AI music generator platforms have grown virally for the revolutionary ability to produce music; but consequently, they are arguably also the default repositories for consumption. How apps organize content has no incumbent mold but tends to fall somewhere along a spectrum between two organizing principles: social graph and interest graph. On one end – say Facebook – content is almost completely socialized. Users primarily consume what their social network shares, even if the content is low-quality. TikTok or Soundcloud – somewhere in the middle – balance social proximity with its best guess on user taste. YouTube – on the other end – almost entirely prioritizes its recommendation engine. 

With generative AI music, our panel seems to prefer the model of the incumbent – that is, Spotify-like organization that prioritizes known artists and recommendation engines above charts and social networks. The most striking takeaway – one that will repeat across the next questions – is the retained deference to creators, even in a world where AI-powered music tools distribute creation.

II. Artistry

 Last week, the estate of Tupac Shakur sent Drake a cease-and-desist letter after the artist used AI to convincingly imitate Tupac’s vocal in “Taylor Made.” Putting aside the legal battle, the song questioned how consumers value artistry – more specifically, if imitation of artistry is close enough to the real thing. Our panel answered a Taylor-Made-inspired hypothetical. Imagine a friend sent a song by their favorite artist. How would their opinion of the song change if they discovered afterwards the song is AI-generated.

Consumer reaction depends. A majority of respondents (67%) would change their opinion of the music track       if the vocals were AI generated; whereas a minority (45%) would change their opinion if it were the instrumentals. While artists and music producers historically create both, vocals are typically more recognizable, more emotional signals of artistry. The difference to us reflects the resilient connection between human artist and fan. Definitions of creation may evolve, but our panel shows the demand to retain – in whatever form – that fundamental relationship.

Platforms like Suno demonstrate exactly that – how artistry can evolve while still maintaining the fundamental connection to its audience. The AI music generator platform explicitly doesn’t produce imitations of other artists (“fake Drakes” in the company’s own words, although “fake Tupacs” might be more current). However, by curating and showcasing its users’ music, Suno enables creators to build AI-native musical audiences. Suno only released in December, but based on our observations on the platform over the past five months, its leading creators get hundreds of thousands of plays.

III. Use case

By our account, AI-generated music platforms productize three distinct consumer capabilities: the ability to make studio-grade music, the ability to personalize existing music or artistry, and the ability to produce otherwise improbable if not impossible musical combinations. While the underlying technology may be similar, these use cases are remarkably different. Our featured song (from our very own Arjun Jain) reflects personal interest in producing and sharing music; whereas, the thousands of videos on TikTok or Reels that leverage AI to produce unlikely remixes or covers (say Squidward singing the Beatles or the presidential candidates in a rap battle) are largely comedic.

Our panel collectively values these use cases equally. That distribution of perceived value reminds me of the stickiest content platforms whose primary value is impossible to distill – platforms like TikTok that have become outlets not just for entertainment, but education, commerce, advice, journalism and so on. New music platforms can similarly be forums for creation, but also for discovery, inspiration, curation, community.

Reach out to abordetsky@nea.com, hworland@nea.com, jkaplan@nea.com to continue the conversation.