Is This Content?

Yes, and Suno is a Cheese Factory

Content. I’ve come to dislike this word. For some reason when people say “content” I always picture it as a plastic tub of cottage cheese. Delicious content, at least the small curd, whole milk variety. Content, without further description, could be anything. But when people use the word to describe art such as songs and albums, I correct them. Content is cheese curds, but art is the lifeblood of culture and our greatest expression of humanity. We all know the difference.

I’ve been railing against using the word content to describe art for well over a decade. Maybe someone else remembers the panel some of us did at Belmont University in 2010, where I accused the tech industry of inventing the use of content to devalue art. I still believe this with all my heart. The implication is that the plumbing has the value, the stuff that flows through is fungible. Anyone who understands the value of music knows this is a fallacy.

I’ve been enjoying seeing Daniel Ek get roasted this week for saying that the cost of content creation is “near zero.” That’s exactly what he said, but the reductive statement wasn’t even the point of the tweet. If you missed it, he stated it as a matter of unquestioned fact as a prelude to some pseudo-intellectual drivel about Marcus Aurelius. Just, like, now that the shit is free, most of it is pointless drivel with no shelf life. But occasionally, it will be remembered. Implicit is that almost everything will be quickly forgotten, and he isn’t wrong. But the head of Spotify saying it carries weight. And you know, “near zero” might be a slightly different figure for a billionaire than for a rank and file musician.

I admit I use the word all the time, albeit through gritted teeth. My work in music often includes digital marketing. You can’t have a conversation about marketing without talking about content. The way most people use the word, however, reflects various categories of marketing materials that support and bring attention to the art. I’m completely fine with that usage. I object to using the word to describe the music we distribute on digital platforms, which generally has a life in other media and in live performance. There’s a hierarchy in that scenario, with humanity coming out on top every time. That’s clear enough, but it doesn’t follow that the low end of that hierarchy is content. Art is art, regardless of the means or the medium. Content is cheese, right? (Yes it is)

The reason I bring it up today that tech innovation has created enormous value for the music business, but it’s also steadily devalued its precious intellectual property. Turned it into cheese, so to speak. I believe we need to talk about this fundamental devaluation because we’re heading into a time when it will grow exponentially. We’re not ready for the disruption generative AI will bring to our beloved creative industries. Now is that time to reassess the value of music, and agree to hold the line when it comes to that value. We need to call out the companies and individuals who seek to devalue our work.

You should know the names Mikey Shulman, Georg Kuscko, Martin Camacho and Keenan Freyberg. These cheese merchants are the founders of the music-generating AI model SUNO, which is likely (though not transparently) trained on huge numbers of copyright protected songs and recordings. They’re all brilliant and highly educated. Suno is amazing - it’s wild that we’re this far along this quickly. Shulman says Suno wants everyone to experience the joy of making music, and that their company wants to achieve this “legally and ethically.” He envisions Suno growing to eventually have a billion people paying $10 a month for the service. That’s more than a third of the present value of the entire recording industry. It is past time to call bullshit.

When people talk about the infringement risks of generative AI models, they often mean the output. Under copyright law, output only infringes if it is “substantially similar” to a protected work. A voice is only infringed under our state-specific Right of Publicity laws if it strongly resembles a famous singer. A tech platform founder recently told me that an AI model should be held to the same standard as a human because a neural network works like a human brain. When we hear music and we’re influenced by what we hear, it’s inspiration, not theft. Our brain doesn’t make copies, therefore it isn’t copyright infringement. That’s true, but AI training is different. I’m far more concerned about the input of copyrighted materials in the training. Until the courts tell us otherwise, we can assume it’s infringement to train a commercial AI model on protected works.

The Suno founders are not from our industry, not part of our industry, and none of them has a professional background that indicates in any way that they give a shit about music or musicians as they claim they do. Maybe they studied music or even played in garage bands, and maybe that would make it worse. They are Ivy League educated academics with deep expertise in machine learning, data analytics, mathematics, and physics. They’ve concluded that a music-generating model is an easy point of entry for an AI startup with an enormous financial upside. Music is low-hanging fruit. Just like Open AI, Suno launched its product without pursuing licenses on the shaky legal ground that training using copyrighted materials might possibly be Fair Use under copyright law. ChatGPT is trained on the entire Internet, as big a training set as possible. Suno is trained on music. I assume it is only trained on good music, with heavily curated data sets. Why would they train it on mediocre or bad music? The product, by the way, is very good. It’s a marriage of a brilliant algorithm that is primarily derived from the work of others and (I believe) a highly curated data set. Take away the data set and it’s just an algorithm. Of course this is all speculation, because Suno hasn’t been transparent about how the model is trained.

I consider launching a product trained on music without license or permission the nadir in this whole “content” scam. It’s the ultimate act of disrespect. They took all the music they wanted, which may have been all the music but is more likely a small sampling of excellent specimens of genre, feel, composition, and execution. They launched a monetized product on a flimsy legal argument, knowing that nobody wants the Supreme Court to weigh in on whether it’s all actually just content. What do you say, Clarence Thomas? His opinion shouldn’t matter, because it should never get to that point. This is between music and tech platforms that use music, and I believe they need us more than we need them. Relying on a legal argument that music has no value is bad faith. Spending investor its hundreds of millions in investor money to push such an argument in court is obscene.

Timing is everything in tech innovation, and Suno’s timing has to be strategic. They’re banking on building a brand on “wow” factor of being the first great product to grow virally. Suno and other models such as Udio are getting ahead of the competition in launching their brand so that everyone else is playing catch-up. The competition will of course be Big Tech. Is the strategy to build a brand and business ahead of the biggies, or is it to be acquired? Open AI, Google, and others will offer products that will likely quickly become better at music generation than Suno. Maybe they’re figuring out how to make deals for the rights to actually launch legally and ethically. Maybe there are already deals in the works, structures and business models being designed. Ultimately, the quality of their products - everyone’s products - depends on access to training sets.

I’m by no means anti-AI. I’m aware that AI models are running in all sorts of applications and creating all sorts of efficiencies in our daily lives. I know that it will transform our lives in very useful ways, as well as stupid and pointless ways. I absolutely do consider the AI-generated song “I Glued My Balls To My Butthole Again” to be useful - perhaps the highest and best use of the tech. I look forward to playing with these entertainment platforms once I am convinced they respect music and musicians. But let’s be clear: Suno isn’t about the magnanimous act of bringing music creation to the masses; it’s about making money.

I believe that the Suno algorithm is useless without copyrighted musical works and sound recordings, full stop. I believe they have shown an unacceptable disrespect for artists, composers, musicians, and rights owners by launching without transparency or payment. I say we oppose any model that doesn’t volunteer to negotiate a structure for licensing and payment. With transparency and open discussion it’s within the realm of possibility to do it right. In fact, it’s a good challenge for an AI algorithm. Come on, machine learning geeks. Figure this out and cut it out with your “move fast and break things” bullshit. Take a seat at the table and be honest about what you’re doing.

I love music so much, it has brought immeasurable joy and value to my life. I love it, even if I can’t stand to listen to most of it. We music people live highly curated lives, curated around what we choose to listen to. People who love music listen actively, and we won’t settle for mediocrity. We won’t settle for content. Maybe we’ll mess around with your music generating platform and marvel at what it can do. But in the end, we listen to music to be moved by authentic expressions of humanity. What if AI models became so good they are able to create with beauty, complexity, emotion, and insight beyond our capability? I say that would suck. That can only happen, I believe, with curated training driven by access to proprietary music assets. I say we don’t want it. Music is the greatest expression of our humanity, just like all forms of art. We should welcome machines to enhance and facilitate our human creativity, not replace it. That’s the mission Suno claims to pursue, and we should hold them to it.

The dystopian future we must actively avoid is one where humans create content to feed the algorithms that create art. That’s unacceptable. The models must serve creators, not the other way around. We’re swimming in a seas of cheese all day on our endless scroll day after day thanks to tech innovation. Are we better off? In some ways, sure. But we’re more disconnected than ever, trying to sort through what’s real and what is some algorithm trying to shake us down. Most of us don’t want to hand over our humanity to tech platforms vying for our slack-jawed fragments of attention with more and more cheese. That’s why we need to protect the value and integrity of human creative expression. All you brilliant scientists out there building the future should be building us tools to make us more connected through art and authentic humanity. And you should do it respectfully, in a way that always acknowledges the value of human creativity.