Water and Music: Thought Leaders on AI in the Music Industry
Note: Our company is not affiliated with Water & Music. We consider them the gold standard for data-driven research and journalism in the AI music field.
Imagine a new species of intelligent machine that studies musicians and mimics their style using a neural network. Would it be a copyright violation to train those AI musicians on protected material? What if the copycat music started cutting into streaming revenue for the artists that it trained on?
The music industry has been known to chase down royalties when a big enough artist samples their music without permission. But the legal infrastructure that's been built to address sampling rights might not be enough to combat generative AI music technology. Machine learning has developed so quickly that new laws are being negotiated and defined in real time.
To better understand the nuances of artificial intelligence and how it could impact the music industry, we've been reading the reports from Water and Music.
What is Water and Music?
Water and Music describes itself as a research and intelligence network for the new music business. In December 2021, they published an inaugural collection of their own research into music and Web3. Each year they explore new topics, from NFTs and the Musical Metaverse to the current season on AI Music.
Their founder and publisher, Cherie Hu, leads a team of research professionals and oversees the Water & Music DAO. An expert in her field, Cherie was recently interviewed for the book Rise of Virtual Communities, alongside a dozen or so tech founders from Reddit, Flickr, Deviant Art, Discord, Second Life, etc.
In February 2023, Cherie published a chart of music industry power dynamics on Twitter, highlighting the influence of global financial groups like BlackRock on music industry giants like Warner Music Group. In 2022, BlackRock carved out a $750M fund to acquire stakes in music copyright that would contribute to their $10Trillion in total global assets.
From this high-level view, it's easier to see how shareholders apply pressure and shape the trajectory of the music industry. When Warner and Spotify launched a major podcast partnership in 2021, were BlackRock and Tencent pulling the strings?
It's up to each of us to do our own research and draw our own conclusions about what's happening here. What your interpretation?
Objections to AI in the music industry
In October 2022 we covered the RIAA's complaints regarding free AI mastering software popping up onilne. But in a plot twist, the developer responded explained they had simply used an open source separator called Spleeter, created and promoted by Deezer, one of the largest legal streaming services in the world and a key partner of several RIAA labels.
Just this week, in April 2023, Universal Music Group began publicly circulating their concerns about AI generated music on streaming services. They claim that neural networks can now study the vocal and instrumental styles of their artists to produce new works, at a pace that may disrupt the industry's earnings.
PluggingAI, a YouTube channel with 2.1K followers as of mid-April, was cited in the Financial Times as an example of an offending party. The channel's most popular content shows famous rappers singing songs that aren't their own.
On April 14, Business Insider reported similarly on an AI cover song that features Rhianna singing a cover of Beyonce's Cuff It. The Twitter user behind the cover is called Rhianna Facts, a dubious name with a description that reads "Don’t miss any Rihanna’s updates again turn on". Each cover song is labeled "IA", an inversion of "AI" gaining popularity among some underground communities.
The Insider article points out that other videos, like this Kanye West deepfake of the Plain White T's song "Hey There Delilah", have amassed over half a million views in the past two weeks, on a channel with less than 3K followers.
Serious impact of vocal deepfakes beyond music
All signs point to AI cover songs as the battle ground where music industry executives will build their argument. The impersonation of a celebrity voice has become easier than ever in the past year, thanks to text-to-speech services like Uberduck. But in most cases, producers needed a pitch shifting app like Melodyne to give the spoken audio a melody. This technical challenge has slowed the growth of celebrity vocal deepfakes, until now.
A new generation of voice transfer technology has emerged this year that will make Uberduck look primitive by comparison. One app, called Voice.AI, is scheduled for release on April 22nd and allows anyone to upload raw vocal data, in order to train on it and imitate natural speech with vocal tone transfer.
Celebrity voices are easy targets, due to the sheer volume of vocal data that exists in the public domain. On April 4th, a new YouTube channel called the Joe Rogan AI Experience published its first episode (shown above).
The real Joe Rogan responded to the situation saying “This is going to get very slippery, kids.”
The next generation of voice transfer tools will create something indistinguishable from reality. That's a serious problem that extends beyond music. Yet music is a multi-trillion dollar industry, so there's no question that vocal impersonations of singers could eventually dilute the content pool of artists and impact their brand.
Comparing AI cover songs to the original music
How similar are these vocal deepfakes to the original? You'll have to take it on a song by song basis. This 2018 hit song Sunflower by Post Malone and Swae Lee. The track was featured on the movie Spiderman: Into the Spiderverse:
Compare it to this AI deepfake of Kanye singing the same words and melodies. You can still hear enough audio artifacts in Kanye's AI voice to know it's an imitation. But how much longer will we enjoy the ability to tell them apart?
Meanwhile, as vocal deepfakes become the bane of the recording industry's existence, other music industry sectors are benefiting greatly from these innovations. Let's return to Water & Music to round out our perspective a bit.
Water & Music's Creative AI Music Software Database
You might be wondering which apps people are using to create these AI cover songs. From source separation to vocal synthesis, there are more than a hundred AI music apps available for experimentation as of 2023. Water & Music currently hosts the most complete, living database of AI music software that we're aware of.
Visit the Water & Music website to read the full reports and get the inside scoop on how these AI tools will impact the music industry.
Below you can find a snapshot of four pieces they've published on the current state of music AI tools. This includes an analysis of AI Sentiment across the music industry, details about AI and music copyright, as well as guides to building a creative AI community.
For those who are interested, Season 1 and 2 focus more on music and NFTs. They offer a unique look at failing music industry models and the value of transitioning to decentralized autonomous organizations. You can read up on Cherie Hu's DAO and the $tream governance token, to better understand how it's used to manage equity amongst shareholders.
We believe our core readers will enjoy Water & Music. If you do subscribe to their journal, be sure to join their private Discord channel. Explore our list of the top ten AI music Discord channel to browse other communities as well.