Living in the Future; Learning from the Past
A Time To Gain, A Time To Lose, A Time To Call Bullshit
Think about how messed up this is. Tech companies are using all of the history of recorded music to train AI tools for commercial music applications - to earn untold BILLIONS - creating original (yet derivative) music with a goal to do it better than humans. Computers…to CREATE music…derived from all of human music creation…BETTER than any human creator that ever lived. This mission is already well underway with impressive models popping up all the time. Sure they suck for now, but they’ll get better. Eventually….with massive data sets and intensive training…they’ll be truly great! Better than Mozart, than Dolly, than Prince, etc. etc.
For those of us who believe that music is primarily an expression of our HUMANITY - which I hope is everyone in the music ecosystem - it’s time to call bullshit on this bullshit. I am not anti-innovation, anti-tech, or even anti-AI. What I’m against is using our collective work - including some of humanity’s greatest achievements - to outsource human creativity to robots. But guess, what? None of it happens without the training data, and we control the data.1 Therefore, we can shape the future. Tech companies are on the wrong mission. They should create tools to unlock and enhance human creativity, not replace it.
I had lunch recently with a big time record executive, a big deal guy who I admire. These days he runs a massive operation, leads a team of hundreds at least. He’s based in L.A., and we were talking about how many Nashville music executives are hopelessly behind the curve, a common opinion held by Coastal execs. He said, “I only want to work with people who live in the future.”
I agreed - he wasn’t looking for an argument. And I do agree - I know what he meant. Living in the future means being ahead of disruption. New technologies have forced change in our business for more than a hundred years. We don’t have a choice about the disruption, though some thrive with the changes while others eat dust. When you read about massive layoffs at major music companies in recent years, know that it probably isn’t due declining revenues. More than likely it’s leadership getting ahead of disruption and letting go of soon-to-be obsolete approaches and ideas - aka living in the future. Harsh but true, and necessary.
When I first worked in music over 20 years ago, I became a full-throated advocate for on-demand as the best path forward for recorded music. I’m proud I saw the future, but I used to regularly piss some people off - people who wanted to protect the status quo - the product-based business. I believed on demand made the most sense when you looked ahead to how mobile was likely to develop.
It’s an oversimplification to say that the tech industry threw music a lifeline twice with the iTunes and the on-demand streaming business, but there’s truth there. The tech industry, however, has shown again and again that its forays into music are more about access to content and less about curation - the advent of that use of the word proves that tech companies don’t really give a shit about music. What’s wrong with calling creative music content?2 When you design structures for digital commerce, you need stuff to flow through the pipes, stuff that people want. Stuff that holds our attention. You can’t charge for the infrastructure without the content. That should mean the music IS the value, but tech world doesn’t see it that way. The systematic devaluation of music didn’t happen by accident. Using music as training data without permission or payment is the next frontier in the devaluation of our work.
As we look to the future, we need to keep in perspective what hasn’t changed throughout all this disruption. Great artists, great voices, great songs, great records that stand the test of time. Live music. Music matters to us because it’s a pure expression of our humanity, a highly evolved way to share our experience. Music evolves because innovators put in the hours to master their craft, dig deep, and share thoughts, ideas, and experiences that resonate. Music evolves because fans support and encourage musicians. Regardless of the technology that facilitates creation and distribution, it’s always been about connection and community.
That is why the next disruption is more of a threat than an opportunity if we don’t take action now to create structures and guardrails. Developers of large language models such as ChatGPT have used the entire Internet to train generative models to emulate human behaviors and processes. The models are incredible, truly a mind-blowing pinnacle of innovation. However, it’s likely they violated all sorts of intellectual property laws in the process. We know these models will be extremely disruptive in positive as well as negative ways, but they’ve led to progress. AI models are deployed in all sorts of helpful, mostly invisible ways. That’s good - tools assist in the creative process through distribution and marketing…they help us create. But what about the creative process itself? That’s where we need to have a serious conversation.
The ultimate goal in AI innovation is AGI (artificial general intelligence), for generative models to do everything better than humans, including creative expression. Algorithms without training data are essentially useless, but with robust training sets they can accomplish incredible things. Whether or not they meet the legal definition under copyright (spoiler: they do), generative AI works are derivative works. I’m very much in favor of innovation that makes our lives easier and better, more productive, enhances the creative process. As a musician, I’m looking forward to incorporating innovative models into my own work. But here’s where I draw the line: these derivative AI tools should enhance, not replace, creativity. Living in the future is fun! But if we cede creativity wholesale to computers, let them have it all without cost, limits, or terms, then music has lost its essential value to humanity. The entire music ecosystem, from the biggest music companies on the planet to the individual hobbyist creators like me, should tell tech companies to piss off until they have a plan to build under the control, direction, and full monetization of the music rights community.
How do we do this? Simple. We enforce existing intellectual property laws and lobby for new ones where the old laws are outdated. We force tech companies through litigation to operate within a licensing and payment structure that requires full transparency regarding all training datasets. No matter how perfect the algorithms become, ultimately they are only as good as their training data. If licenses are required, we are part of the conversation about how our work is used to train models, and what the objectives are. Tech companies scream fair use, but that argument is conceptual and not at all settled under our laws. They will use the courts, so our deepest pockets must dig a trench and fight. AI will never challenge human creativity without full access to all of the music that provides the tools for our creative expression. From large platforms to individual creators, we must have the option to say no to bad ideas, and to insist upon compensation for models that serve our interests. We have the law, the courts, and if it’s an industry initiative, and the funds are there if we’re aligned. Sometimes it’s easy to be on the right side of history.
That is, outside the public domain. They should go ahead and gobble up all that stuff under Creative Commons licenses. All those “information wants to be free” folks can suck it.
I’ve made peace with the word content to describe a certain type of media - never call someone’s song, novel, movie, or record album content. Content is all the stuff we make to market and bring attention to the art. It’s social media posts, bios,
Could we use AI (or an actual human) to file a copyright infringement case for every one of the 100+ songs I have out there on Spotify/itunes right now? Since surely these large data sets are using them right? I mean it would cost them so much to defend every suit.
To me, the basic fact that machines/algorithms can't empathize, they can only manipulate emotions, is the big dividing line that holds out hope for creative human expression. Even among us old obsolete human beings, there are people who approach songwriting like a math problem or carpentry project, vs. those who seek to emotionally resonate with others.