The Current State of Creative AI: What Songwriters Need to Know

The rise of creative AI has some songwriters and publishers understandably fearing for their future. Will AI-generated music ultimately compete with artists and songwriters for real estate — and revenue — on terrestrial radio and streaming playlists? Will a fully AI-driven culture devalue human creativity altogether?

While some areas of recorded music, like productions, may be under threat, creative AI has largely been a positive force for songwriters and publishers who view it not as a competitor, but as a tool to augment otherwise difficult creative processes. In fact, creative AI might have the same long-term effect on society as the first drum samplers and digital audio workspaces (DAWs): democratizing music creation and expanding the market of amateur and professional artists, which ultimately benefits the broader music industry.

Here are four main takeaways about the current state of creative AI, and what it means for the songwriting and publishing community:

The Majority of Investment in Creative-AI Experimentation Is Coming Not From the Music Industry, But From Tech Corporations and Venture Capital

Big-tech companies and VC firms see significant market potential in creative AI applications and are pouring money into R&D that inventive creators can follow.

Google has been building machine-learning creative tools for artists through its Magenta research project, while Facebook’s Artificial Intelligence Research (FAIR) team built an AI system (on top of Google/Alphabet’s own WaveNet model) that can autonomously translate audio recordings from one instrument, genre, or musical style to another. Spotify is also currently developing AI-powered tools for artists under its Creator Technology Research Lab, based in Paris.

Aside from these big-tech behemoths, a growing number of startups (e.g., Amper, Popgun, Jukedeck, and Amadeus Code) are building their own AI-powered composition and songwriting tools. This productizes cutting-edge tech not just for artists, but also for a mass market of music fans and consumers. Financially, these startups have also collectively raised millions of dollars of VC funding and institutional investment as alumni of music startup accelerators like Techstars Music (which, notably, has backing from major record labels). 

Indie Artists and Songwriters Are Already Using Creative AI in Their Everyday Processes — And Getting More Creatively Inspired and Enriched, Not Less

Several of the above-mentioned startups have partnered with artists on groundbreaking new musical projects. For example, singer-songwriter Taryn Southern co-wrote an entire album (I Am AI) with the help of four creative-AI programs, including Amper, AIVA, Google’s Magenta, and IBM’s Watson Beat. Canadian songwriter David Usher co-founded AI creative studio Reimagine AI, and has worked with Google Brain to develop an automated lyric assistant called Lyric AI that premiered at major computer science and tech conferences. Beatboxer and Nokia Bell Labs artist-in-residence Reeps One is working with engineers to develop a beatboxing AI that draws from and expands upon Reeps One’s own style and techniques, and can even battle head-to-head with its human origin onstage.

Some AI-composed songs are even being recognized on select Spotify playlists. Under the artist name SKYGGE, French songwriter Benoît Carré used Sony’s Flow Machines AI to help compose every single track on his album Hello World, bringing onboard renowned human artists like Kiesza and Stromae to help bring the compositions to life. The lead single from the album, “Hello Shadow,” made it onto Spotify’s flagship New Music Friday playlist last year, as well as on localized NMF playlists in the U.K., Norway, and Scandinavia.

Not coincidentally, François Pachet — the former head of Sony’s Computer Science Laboratory in Paris, where Flow-Machines was developed — now heads Spotify’s Creator Technology Research Lab.

The Production Music and Sync-Licensing Sectors Might Be the Most Threatened by Creative AI…

Production music (i.e., recorded music tailored for licensing to film, TV, and other media), is still a major revenue stream for many labels around the world. For instance, Universal Production Music, based primarily in Europe, releases over 45 albums a month with offices in over 20 countries.

The rise of AI, however, arguably puts production music the most at risk. Consider how the target market for startups like Amper and Jukedeck are “content creators” at large — e.g., burgeoning filmmakers, YouTube personalities or Twitch live-streamers who don’t want to deal with copyright infringement issues but might not necessarily have the budget to purchase or license background music properly.

In addition, consider the way most sync-licensing transactions work today: music supervisors for a movie or TV show send a brief to as many as 40 publishers at a time, detailing exactly what parameters they need for a song (mood, tempo, instrumentation, etc.), then each publisher sends back at least a dozen tracks, putting the burden back on the supervisor to choose just one song from that entire pool. The process is high volume and high stress, with lots of manual back-and-forth communication.

It’s not far-fetched to imagine an AI cutting this entire process in half by automatically and satisfactorily generating a piece of music based on the rules outlined in a particular sync brief. While AI has a long way to go before it can meaningfully compete with humans in a creative context, it can beat humans at low-cost output in record time based on a predefined set of rules (see: Google Translate or AlphaGo as examples beyond music).

…But Most Artists Shouldn’t Be Worried About Losing Their Careers

From a legal perspective, one of the most heated debates in creative AI revolves around whether an AI or algorithm can own copyright and generate revenue for itself. As of today, creative AI is still in such early stages of development that officially granting AI “ownership” over a track happens only on a case-by-case basis.

For instance, AIVA, one of Southern’s AI-powered collaborators, is officially registered under SACEM (the first-ever algorithm to have such status). Southern also chose to credit Amper as a co-writer on her single “Break Free.” In contrast, Usher uses Lyric AI as an “assistant,” but not as an officially credited co-writer. Likewise, with Hello World, Benoît Carré is credited as the main songwriter under the SKYGGE moniker, with no royalties going to Sony or Flow Machines.

Furthermore, Southern and Usher are clearly the primary writers of their own songs, even if they use an AI assistant to expedite the process. In each of their cases, the artist is ultimately the entity with the last word — the person who defines the parameters under which the AI searches and creates content, and who then curates, edits, and exercises judgment over the output of the AI’s work.

The ability of AI to “compete creatively” with humans is also arguably limited by our own understanding of how we work. We cannot create an algorithm for “good,” “beautiful,” or “creative” music if we struggle to define the rules for creativity ourselves in the first place.

As previously discussed, the areas of the music industry where we already consciously put these rules in place — production music and sync licensing/supervision — will have to reflect on and revamp their value proposition in an AI-driven future. Nonetheless, songwriters and publishers should still approach creative AI today not with fear of displacement, but with a sense of possibility to break new creative ground that was literally undiscoverable with previous technology.

As long as you own a share, or are the rights owner to your composition, Songtrust can help ensure you're affiliated with the right collection organizations and that you're collecting all of your mechanical and performance publishing royalties globally. Don't leave money on the table; register today!