Updated: Jan 20
What you’ll learn in this article:
Now and then, a new technology or trend in music production gets people thinking “music is dead” and “real” musicians aren’t needed anymore. Yet, from samplers and digital production software to Pro Tools and Autotune, there are still singers, songwriters, producers and engineers making music, and the industry is very much alive.
But what happens when we have a technology so powerful that it can literally replace all the creators involved in the process of making music? That’s the dilemma musicians, publishers and record labels are dealing with, concerning Artificial Intelligence (AI), as it has evolved to the point that it can create entire songs or tracks, with the only human involvement being in telling the AI to do it.
Pixabay, via Pexels.
There are levels to how AI can help a songwriter or producer. For one, AI can help as little (or as much, since doing this manually can be a nightmare) as in source separation — that is, isolating individual sounds or groups of sounds in a track, such as Spleeter, an AI tool powered by music streaming company Deezer. From this perspective, AI can boost music production by giving creators access to pre-made pieces of arrangements, facilitating sampling and leaving the creators’ work to be one of curation more than a technical one.
But more than just offering ready-made elements, AI can also help bring new ideas to life. “Not everyone works on all the stages of music production,” Brazilian music producer Numa tells SPARWK. “Sometimes I can imagine what I want for a song, but I’m not able to materialize it. So if AI can bring out elements that a person can’t create, then that’s great.”
Composition-wise, AI tools such as lyric generators can help overcome writer's block or to help lyricists or songwriters that lack a collaborator with complementing skills. An AI-written lyric lacks feel (and sometimes, lacks sense too,) but it’s a relatively harmless implement, at least in comparison to other things AI can do.
Things get more complicated when we have AI composing and producing entirely new songs, vocals included — and these could even be vocals from dead singers. Attempts at creating songs by Elvis Presley and Nirvana have gone unsettingly well, and projects like Holly Herndon and Jlin’s “Godmother” (which features the AI “artist” Spawn) are using AI to reimagine artworks through the voice of lost loved ones.
It feels like any individual can do anything with AI. But are they allowed to? And most importantly: do people want to listen to music made by AI? These are the questions that will define how much AI will impact the music industry.
Where AI music gets tricky: Intellectual Property
It’s all fun and possibilities until the discussion regarding AI in music hits a block: Intellectual Property Law. If incorporating AI into the music-making process is meant to save time and money, it’s only right that a label wouldn’t want to risk said time and money by getting copyright lawsuits. However, IP Law might not be prepared yet to deal with all the quandaries that AI music brings to business.
The discussion is not new: the US Copyright Office (USCO) has been reporting on the “Problems Arising from Computer Technology” for as long as 1965. Almost 60 years later, in September 2022, the USCO granted what is allegedly the first copyright registration for an AI work — but less than a month later, it showed signs that it will backpedal due to the lack of “human authorship” in the work.
The main issue with AI-made music relies on copyright ownership. According to Dr. Luca Schirru, an Intellectual Property researcher who specialized in Artificial Intelligence for his Doctoral Degree, assuming AI can create an objectively original work doesn’t necessarily lead to it being deserving of copyright protection.
While copyright systems may vary (in countries such as France and Brazil, for example, copyright law is intrinsically tied to personal rights,) the human intellect remains a key element for a work to be considered worthy of copyright protection. An AI tool can “write songs,” but it cannot be considered a music rights holder simply because it’s not human. Entertaining the idea of an AI tool as a creator whatsoever wouldn’t take away the risk of individuals claiming rights over AI music either (there’s debate on whether AI developers or users can be copyright owners.)
Even for the smaller uses of AI, such as in source separation libraries, there can be problematic outcomes. Pulling out different elements from different tracks can create what copyright law calls “derivative work;” however, these require permission from the copyright owner (unless the song and/or the record is in the public domain), and many creators don’t bother getting these. That’s far from the first time technology has facilitated copyright infringement, but the harms that these practices help foster are a part of the discussion around the ethical implications of AI use in music.
No less complicated is the idea of plagiarism between AI music and human-made music. But some think AI actually can be used to avoid plagiarism disputes. Musicians Noah Rubin and Damien Riehl (who’s also a copyright lawyer) developed an algorithm fed with allegedly “all the melodies” possibly created so that these can be copyrighted and used by musicians without risk of copyright infringement. Whether an 8-note combination is subject to copyright protection is highly debatable (the Court that ruled Katy Perry’s appeal against rapper Marcus Gray in “Dark Horse” believes not), but while all these questions are pending answers, the insertion of AI in the creation process fosters a grey area that labels might not want to put money on.
Can AI music ever be as good as “human” music?
With the risk of AI making creators go obsolete, you would think AI would not be welcome by the songwriting community. However, not everyone is threatened by it.
“If AI can learn different plug-ins for various types of songs, genres and styles of certain songwriters’ and if I can have these customised settings with a touch of a button, it would greatly save time for many songwriters,” Korean songwriter Houdini told SPARWK. “And if songwriters can create great music with AI assistance, that would be another milestone for humanity in music.”
To think of AI as an assisting technology for the songwriting and production process is to think of people being able to make more music, also in easier and quicker ways. One could wonder if this is something anything wants at all (don’t we have enough music already?); but also, one could wonder to what extent it would preserve the “human element” of music.
Fundamentally, what music AI tools do is not much different from what a human creator does — that is, pulling out from the vast yet limited repertoire of notes, chords, and words, and combining them in a way that makes an original song. Melodies may be “just math,” as said by Damien Riehl on Ted x Talks, and in that sense, an AI’s “songwriting skills” can only be as good as its ability to make calculus, as its user's ability to command it. In an interview with Rolling Stone, Sean O’Connor, one of the directors of the project that created a “Nirvana” song, said 90% of what the AI generated for them was “really bad and unlistenable.” “So you start listening through and just finding little moments that are interesting.”
What we have yet to see is how good AI can be in linking up pieces of data to generate music that can compete with human-made music. Given how passively music has been consumed lately, it’s not at all unthinkable that someone may enjoy AI music in the same way that they enjoy music made by an individual. “Consumers don’t often think about if this music is made by AI or a human,” says Houdini.
It’s not hard to envision people liking an AI-generated song such as “Drowned in the Sun” as much as a song written and recorded by Nirvana themselves. “[‘Drowned in the Sun’] is accurate enough to give you that [Nirvana] vibe,” singer Eric Hogan, who participated in the AI project, said to Billboard. Hogan’s choice of words is suggestive: AI music may be more about the “vibe” than the actual efforts or emotions of an individual. And ultimately, just the vibe may be enough for many; but not for everyone.
People may enjoy listening to “Drowned in the Sun,” but would it satisfy a hardcore Nirvana fan’s desire for access to material from their lost idols? There’s a deeply personal bond between fans and idols that an AI could not recreate or replace. Fandoms aside, human songwriting may differ from AI songwriting in how it not only “combines” lyrical and melodic elements, but also in how it does it in ways that are meant to convey specific sensations, or even reflect deeply personal experiences.
If “creativity is always at play when concrete things are perceived to mean something that goes beyond their practical purposes” (as said by Jungian analyst Dr. Paul Brutsche, in Creativity: Patterns of Creative Imagination as Seen Through Art,) then it’s the ability to make these non-practical connections that AI will always be missing — although it’s worth noting that songwriting can also be a highly strategic task, and not just AIs, but humans can be trained to convey sensations through music too.
Photo credit: Tara Winstead
AI will replace a few skills, but will require some new ones too
There’s no going back to using AI in the music industry. But besides the sound of music, AI can also change the way the industry is organized and the type of expertise it will require from professionals from a wide array of industries (music, tech, business, marketing.) Justin Hahn, Chief Creative Officer of Joombas Music Group, believes AI can impact more fields than just music production: “Everything from the structure of music rights to the essence of the actual art embedded into them are dynamic and complex, so I’m personally curious to see how A.I will fit into the framework of that”, he shares with SPARWK.
Paraphrasing singer-songwriter Grimes, we may be testifying “the end of art – human art.” Applying that mindset to business, with every skill replaced comes the need for a new skill, and the growing use of AI will lead to a demand for developers but also specialized producers and even A&R’s and executives.
Beatmakers and DAW user producers are tech-savvy by design, but AI can introduce a new type of producer, with highly specialized skills in AI manipulation. Translating consumer preferences or music business strategies into AI commands can also become a skill in demand. After all, the music of the future may not be made by humans, but at the end of the day, it will still be made for humans (even if it’s not a stretch to think of AI music aimed at other living beings too.) Making better AI music that can successfully engage people or cater to their needs will be a challenge involving many links in the industry chain besides just the AI developers.
And of course, we can’t forget the Web3, the metaverse, and all the changing realities through which AI music can be created, consumed, and shared, generating new opportunities and challenges for technology and Intellectual Property as AI and human life coexist and act together.
* * *
Do you think AI music will ever be as widespread as music made by humans?
The world is changing and we want to make it better for music creators.
Join SPARWK us as we work towards a brighter future for all the stakeholders in the music ecosystem.
This content is published for informational purposes and does not constitute legal or financial advice, or promotion of any of the companies or artists here mentioned.