✘ Musicians harnessing AI, while doing no harm
And: Mysterious marketing and gaming the algo; The Stanley tumbler; Crypto's three body problem
In the past year, we've seen countless sensational headlines about artists publicly exploring AI in loud, revolutionary ways. For example, we saw Grimes and Holly Herndon democratizing the use of their deepfaked voices or crafting entire albums using AI generated vocals. We also all heard the stories about Drake, The Weeknd, and Bad Bunny fighting to take down viral songs published using their voices without consent.
Stories like these may lead the average creative to believe that exploring the use of AI in music has to be an extreme and polarizing decision, fraught with risk and the potential for doing harm to other creatives, but the reality is that musicians have quietly been interacting with helpful AI and machine learning for years, often without realizing it. Think T-Pain with his Autotune, sound engineers making automated sound corrections, and personalized playlists on Spotify.
Rather than simply compiling another overwhelming list of newfangled AI music startups, let's delve into real world use cases that well-known musicians have experimented with to integrate AI into their music production and distribution flows in recent years - while doing no harm.
The Beatles and Darkchild: Getting stems
Obtaining high quality "stems" (isolated audio tracks containing a single element of a song, such as just the drums, voice, or piano) is a simple yet powerful use case for AI in music.
The Beatles: “Now and Then” (1977 - 2023)
The story of how The Beatles made their “last” song, “Now and Then,” centers on AI enabling them to pull a stem of John Lennon’s voice, recorded on a cassette tape in 1977. In 1995, Paul, Ringo and George tried to finish the song, but couldn’t separate John’s voice from a low quality audio pairing of him playing the piano, so they abandoned the project. But in 2022, during the making of their documentary "Get Back," producer Peter Jackson created AI software named “MAL” that could isolate all the elements of a song as stems. This allowed them to finally remove the piano from John’s original demo track, and the remaining Beatles could layer on new drum, bass, piano, slide guitar, and string arrangement to create a completely new, genuine Beatles song, almost 30 years later.
Darkchild: SZA’s “Forgiveless (ft. Ol’ Dirty Bastard)” (2022)
Similarly, producer Rodney Jerkins, also known by his stage name Darkchild, used AI platform Audioshake to pull audio of Wu-Tang Clan’s Ol’ Dirty Bastard off of a long forgotten, 25-year old VHS tape of the late rapper (with consent from his estate), and sampled it for SZA’s 2022 track “Forgiveless.”
Weezer: Writing alternate lyrics
One of the best known use cases for AI chatbots like ChatGPT and Bard is copywriting. It stands to reason that songwriters can riff with AI to break out of writer’s block, come up with variations on their own lyrics, come up with song titles and more.
Weezer, “I Want a Dog” (2022)
In late 2022, Rivers Cuomo of Weezer announced the release of a new song, "I Want a Dog," simultaneously revealing that he had used ChatGPT to write an alternate version of the song lyrics. He published them both and challenged fans to guess which lyrics were crafted by the AI. The 2 versions are fairly on par until about halfway through, when Rivers’ lyrics take a much “angstier” turn, a dead giveaway according to his fans, who could mostly tell human from robot.
Evan Bogart: Creating better song pitches
Songwriters have long employed soundalike demo singers to perform their songs when they are pitching songs to specific artists. This can now be done via AI.
Evan Bogart is a top songwriter for artists like Rihanna and Beyoncé, who also runs his own publishing company and label, Seeker Music. He talks openly about how he encourages songwriters on his roster to use AI to create “pitch records,” which are songs written by those songwriters, performed as demos. The AI platforms they use can transform those pre-recorded vocals, usually sung by the songwriters themselves or demo singers, into the voice and tone of the actual artists to whom they are pitching. Talk about personalized sales.
Björk: Creating ever changing, context-aware music
Here’s where things start to get a little crazy. A less talked about, but boundary-pushing use case is turning a traditional “static” song into a living, breathing work that can respond to environmental context, its audience, and take on the form of an ever-changing performance.
Björk: “Korsafn” (2020)
Björk, in partnership with Microsoft, created an everchanging, infinite choral piece generated by AI algorithms that study visual patterns of clouds, sunrise, sunset, birds, and planes from a rooftop camera installed in the Sister City Hotel in NYC. The camera captures the activity of the sky 24/7. Artificial intelligence then transforms these physical elements into choral sounds created by Björk herself over 20 years ago. The resulting notes find their way into the lobby of Sister City as a choral performance.
Disclosure: Translating music into a visual medium
The real transformative power of AI begins to emerge as we consider how it gives musicians the ability to reimagine or pair their music with other forms of art (in real-time or not), and become not just musicians, but visual artists, performance artists, and so much more.
Disclosure: “Simply Won’t Do,” Bronze Editions (2023)
Late last year, Disclosure announced a new project in which it created 1000 unique versions of its song “Simply Won’t Do” with accompanying custom artworks in partnership with new AI music platform Bronze. An art collective, Grotesk.Group, then transformed each unique song version into 3D data points, forming the basis for beautiful visual renderings of each song, which are being sold as 1-of-1 NFTs.
Warner Music x Edith Piaf: Bringing life to a deceased artist
While there has been much discussion, rightfully so, around ethical concerns about creating posthumous music in the voice and sound of deceased artists, there are other ways of using this new technology to celebrate and honor artists who have passed.
For example, Warner Music Group is creating a full-length animated film about the life of the late French singer Edith Piaf, partnering with production company Seriously Happy and her family’s estate. They’ll be using AI to simulate her voice to narrate her story and generate visuals of her throughout the film. According to WMG, their AI has been trained on hundreds of Piaf’s voice clips and images — some of which are more than 80 years old!
In conclusion
“The night is dark and full of terrors” may be a common perspective on the AI revolution for many musicians. And it’s certainly true that we will be battling existential concerns such as algorithmic biases and royalty distribution challenges for AI generated music, a potential loss of human connection in such music, job displacements, and the carbon footprint of AI.
However, it is also possible to be an artist navigating these waters in a way that is beneficial to creative society, while also being revolutionary for the world of art and music. These examples are helping to light the way.
About the author
Lennie Zhu is the Founder and Head of Product at 01 Labs, a product management consultancy that specializes in building software for the creator economy. With almost a decade of experience in product and strategy, Lennie has built products for and advised 10+ top creator startups like Chartmetric, HIFI (acquired by Block), F*** You Pay Me, building for the teams of top creators and media companies such as Billie Eilish, Lizzo, Lil Nas X, Universal Music Group, Mr. Beast, and many more.
If you’re interested in working with Lennie via 01 Labs or hiring her for full-time product roles, you can reach her on LinkedIn or email her at l@lennie.io.
LINKS (via Maarten)
Pariah Carey part three: mysterious marketing and trying to game the algo (MacEagon Voyce)
“Confirming that preparatory work is done is a good exercise in the interim. And as we build and test our content strategies, and release music onto streaming services, remember there are tools like these designed to cut through the noise. They’ll be here when we’re ready.”
✘ Keagon continues with his amazing write up of the Water & Music x Music Tomorrow bootcamp on Music Marketing Data. Here, he contemplates how tools need to catch up, while thinking about how to live within the current streaming paradigm. At the same time, how can you break free from that paradigm?
Why the Stanley tumbler may be half full (Tatiana Cirisano)
“This does not mean that we should not boost music niches, or that going big always equates with “selling out”. Rather, it is important to recognise this tension and find the right balance for an artist and / or scene community, so that you do not end up reaping short-term rewards over long-term sustainability. It is possible that niche scenes are simply diametrically opposed to the sort of success that commercial entities want to extract from them — meaning that labels going after the niche strategy will need to fundamentally shift the way they operate, their definitions of success, and so on.”
✘ I love a good argument for niche over scale, and Tatiana provides a wonderful one here.
Crypto's Three Body Problem (Laura Lotti, Sam Hart, Toby Shorin)
“Censorship-resistant immutability to the law is crypto’s greatest achievement as well as its greatest weakness. By resisting the all-encompassing reach of the law, it creates the conditions in crypto for a new kind of realpolitik, a space where power operates by different rules. However, in removing the law, crypto protocols are left with a three-body problem. Social norms, markets, and code each have their own regulatory logic, often finding themselves in conflict. In this novel game board, protocol designers’ intentions can be undermined, leading to undesirable institutional behaviors, moral dilemmas, and contradictory governance policies.”
✘ I hope everyone will read this. It asks us how we can think about a new model in the terms that we know. This has happened to crypto a lot, and music suffers from this tendency especially. I’ll write more on this, but we need more of this thinking to be able to actually move forward.