I promised when I started this Substack that I would writed about music and generative AI. It’s a big part of WHY I started it in the first place in early 2023. I was (and still am) on a steep learning curve to get a sense of where things are headed, what legal, social, creative, and logistical issues are likely to arise, and how to get ahead of problems. I’ve written a lot but shared very little - in part because in addition to learning about the tech I also had to figure out how I felt about everything. And I’ve felt…conflicted.
My AI Dilemma is this: My entire professional life has been about (at least trying to) facilitate creative people and protect their rights. In that sense, I am deeply concerned about where we’re going and how the development community is forging ahead with technologies that obviously and harmfully infringe the rights of creators and rights holders. On the other hand, I’m an obsessive music fan and creator, and I’m bursting with excitement to be part of the revolution of creation that’s ahead. I love using tools that enable me to be creative, and I strongly believe that a variety generative AI models that are emerging will actually enhance and facilitate creativity. On the one hand, there will be an ocean of mediocrity built on the backs of creative people actually competing for attention and money with authentic creators. But on the other hand, some of the most creative artists will experience a transformation of their work with the tools of the near future.
I’m a music business professional, but at this point I’m also an amateur musician and hobbyist. I recently finished an album I’ve been working on for a decade with the help of some brilliant musicians and technicians I’m lucky to have gotten to know through my professional life. It’s been fairly easy to make music that is technically up to my high standards - both because of my gifted collaborators, but also because of the technologies we already take for granted. I recorded in a friend’s garage on a bare-bones ProTools setup with a few microphones and minimal outboard processing. Gregory Lattimer, the producer, works primarily “in the box,” using digital tools within the ProTools ecosystem. The mix and mastering engineer, the brilliant Paul Mahern, with whom I’ve worked in many expensive professional studios since the 1980s, now does his work in a room in his house. None of this would have been possible 20 years ago. I could have made the record I made, but it would have been far more difficult and expensive to create.
Making the album has made me want to keep going, and to do more things on my own without the assistance of professionals. I’m imaginging what DIY music creation will be like in five years. Here’s a scenario I believe will be commonplace:
It’s the weekend and I want to create and release a track I’ll record a song I’ve already written. Songwriting is the priority for me, so I’ve done that part like I hope I always will: I sat down with the guitar or at the piano, wrote a piece of music and melody, and composed lyrics without the help of a Large Language Model (LLM). I open my workstation and I give a series of vocal commands. I start with a drum set track. I describe the form and feel, and I get very specific about it. I want a 60s sound, a small room with lots of wood, a small drum kit with brushes on the snare but a dynamic performance with a feel like Hal Blaine. I’ll add a bass guitar that’s a finger-plucked hollow body withe flatwound strings, a mic’d amp. My workstation will create something that generally fits my description, then I’ll tweak it until it’s exactly what I want.
Then I’ll describe some additional intrumentation
: a finger-picked acoustic guitar, some light touches from a 60s Wurlitzer electric piano, a gritty shaker with a loose groove, some subtle string pads under the chorus, a baritone electric guitar with a slow tremolo. Once the track is right, I add my own vocal part without much concern about the performance - just has to be close enough to the melody. I have a plug-in that is a model of Dionne Warwick’s voice that I use, with a male harmony from David Crosby. I’ll manipulate the vocals so that they aren’t obviously derived from these famous singers while maintaining the essential qualities. Once the vocals are in place, I’ll describe the mix I want, and then I’ll go ahead and give the command to master the track for commercial release according to my specifications.
When I upload the track for aggregation across digital platforms, I use an LLM to come up with marketing keywords to assist the various platform algorithms to push the music out to listeners. I’ve created several alias accounts on DSPs and social media to make it appear that they are all real artists, and I’ll use generative programs to create short-form video clips to promote the music. It takes me a few hours to complete this entire process.
The first important question is: Is this a creative process? I think so. We don’t question whether a track is creative work when a producer uses drum machines and sequencers. We don’t have an issue when singers use robotic auto-tune settings as a creative choice. Using generative tools with voice commands to create is creative - BUT there are other questions to consider.
In my hypothetical, I released my track on platforms and, presumably, if the track gets any traction I’ll make a little bit of money. It’s possible I’ll make a lot of money! I think it’s fine to give people tools to create in this manner, but everything changes once it’s distributed as a commercial product. None of this is possible with extensive datasets consisting of protected works and named artists. If my command is “make it sound like Hal Blaine,” or “use Dionne Warwick’s voice,” then they (or their estate) should absolutely be compensated under a licensing scheme. If a music generating model is trained on copyrighted materials, those should be licensed and compensated - and rights holders should have the opportunity to opt out of permitting developers from using their work.
As a creative individual, I love the idea that I could create something that’s high quality and amazing sounding because it’s derived from great, iconic performances. There is a creative aspect to me throwing elements together in this manner. But, in the end, it isn’t really MY work. It’s a sort of collaboration with virtual representations of the work of others. The huge challenge ahead is creating an entire universe of permissions and micro-payments so that these collaborations are allowed and monetized so that if I make something of value that people are willing to direct their attention, that my compensation stands behind compensation for the individual creators and owner/rights holders that have made my creativity possible.
I honsetly don’t know if what I’ve described is how it will work for home creators, but it’s reflective of what developers are striving towards. The tech community has a history of disrespect for the rights and value of creators because these things stand in the way of innovation. Information wants to be free, right? I am very much in favor of innovation, and I look forward to having these amazing tools to create, if only for my own enjoyment. However, the foundation MUST support these permissions and payments in a way that values and respects the works that make up these datasets.
As a side note, I’m more and more convinced that emerging technologies will give rise to a renewed interest in authentic human creativity. I’ll continue to use tools that enhance my own creativity; however, I seriously doubt if I’ll ever make any of my own work public that doesn’t reflect my authentic human experience, just as I doubt I’ll ever release music anonymously. Music is fun for me, but sharing music is about communication and community, and just making something that creates a connection with others. I’ve known a lot of creatives, and this is consistent with how most artists feel about their own work. I think it’s possible to keep true to this vision while embracing generative tools, BUT I think most creators will agree: it should be done in a way that respects, acknowledges, and (most importantly) compensates the people upon whose shoulders we stand.
I've long been perplexed at DIY types who are technology-averse and eschew the internet. Just like you said, these are tools of autonomy, both creatively and financially, for artists. It seems to dovetail with that other weird post-punk ethos I've never understood that anything that sounds too good or polished is somehow inauthentic, as opposed to just being the result of hard work, wanting to get it right, and wanting your work to grab people's attention in a positive way.
Are there parallels between this and the business aspects of sampling in music?