How working musicians can beat AI in the marketplace

I’m going to explain how musicians, podcasters, authors and other artists can survive and even thrive amidst the unstoppable flood of AI we will be forced to wade though for the foreseeable future.

Artists, don’t kid yourself, generative AI is here to stay. There is no going back.

But there is a way forward.

This is a personal topic for me. I used to be a professional musician. I put myself though college playing in bars and clubs. I was Japan’s first professional podcaster. I also love generative AI and am excited about the amazing creative potential it promises.

I want to see all of these things thrive. AI will be fine, of course. It’s supported with practically unlimited funds and by lawmakers and industry leaders around the world.

Artists, however, could use a little help.

What exactly does AI create?

People asking if AI can create real art are asking the wrong question. Artists who need to put food on the table need to be asking what artistic needs AI meets in our economy.

With those parameters, let’s look at what exactly AI is creating, using podcasts as an example.

Google NotebookLM can take any textual input (your website’s FAQs, a press release, last quarter’s sales reports, anything) and create a convincing podcast from that input.

A male and a female voice will smoothly and professionally banter about the topic and tease the listener that they won’t believe what’s coming up, and they express broadcast-caliber levels of surprise and admiration over the most trivial bits of information.

It’s really good. NotebookLM has very high production standards.

But there is nothing really inside. After a minute or two, it’s just not that interesting to listen to — even when the input information was interesting.

This is because NotebookLM is incredibly good at imitating the structure and affect of a quality podcast. This is how all LLMs generate art, music, and video. They imitate a particular structure and affect, but the quality of the content is irrelevant.

Structure and affect are the logical and emotional cues that let us classify a work as a particular type of art.

The structure is the logical parameters; a pop song should be about three minutes long, it should have an identifiable melody. An image should be rectangular. An email should start with a greeting and end with a signature. Those kinds of things.

The affect is the emotional parameters. It refers to the emotional reaction we have to a given work. It’s the vibe. Rock and country covers of the same song will have a different affect. They will feel different.

Generative AI is successful today in those areas where structure and affect are important but quality is irrelevant.

Saying “quality is irrelevant” is not an insult or a backhanded way of saying that quality is poor.

The key fact is that AI-generated art (whether it is of high or low quality) excels in situations where quality is irrelevant, and human-generated art (whether it is of high or low quality) excels in situations where quality is relevant.

If you are an artist arguing about the quality of AI-generated art, you are having the wrong conversation.

What is art to us?

Since your future income depends on leveraging this, let’s take a closer look at how we humans interact with art, whether generated by human or machine.

AI music startups like Suno and Udio have announced that they have created hundreds of thousands of “top-40 quality” songs that include vocals and full orchestration. It’s unquestionably impressive, but if you were to ask them “Pick the best two or three, and let’s listen to them together.” it would be an utterly bizarre request.

Quality is irrelevant. There is no “best”. When Suno says they create “top-40 quality” songs they mean they can reliably produce the structure and affect of a top-40 song. All of the songs can plausibly pass for a real song. The quality of any individual song is irrelevant. The quality is not necessarily bad, it’s just not relevant.

For a human counter-example, let’s say your young friend has just put together their first garage band. They play their three best songs for you and ask what you think. You are not going to respond in terms of structure and affect. Saying something like “Wow. That sounds like a real song. I couldn’t even tell it was written by you.” would be so odd it would not even be insulting.

No, you would talk about their music in terms of quality. You might say things like “The lyrics are really clever.”, “The second song had a great groove.”, “You know, maybe you don’t really need a two-minute drum solo.” It’s not that the quality is good, it’s their first band so they are probably terrible. However, we interact with human-generated art in terms of quality. Structure and affect are mostly assumed.

Why things got so hard for artists

The real problem artists face today is that almost all commercial uses of art focus solely on structure and affect, with quality being irrelevant.

As tempting as it is to blame technology, AI did not create this rift between structure and affect on one hand and quality on they other. AI is taking advantage of culture-wide shift in how we engage with art that has been growing for about 25 years.

Unless artists stand up for themselves, it’s going to get much worse.

It’s important to understand that we all used to experience music completely differently. As a kid growing up in the 80s, if you wanted to listen to a particular song, you had to go buy the record. In addition to the friction of going to a physical store, albums cost about $10 and CDs about $15.

In today’s dollars that’s $29 for a record and $44 for a CD. For a student, that’s about the price of a six month subscription to Spotify premium. For a single album!

I’m not saying that was a better system. It was terrible in many ways. However, it forced us to interact with music in terms of quality.

After making such a major investment, you didn’t just put that album on your shelf. You read the liner notes, and you listened to every track multiple times. You then got together with your friends and listened to these albums together.

You would each play your favorite tracks from the album you brought over, say what you liked about it, and everyone would talk about the music — and about all kinds of other things too.

This dynamic forced you to engage in terms of quality. Nobody cared how many albums you had. They cared about what you brought that day and the specific things you liked or disliked about it. “Listen to what he’s doing with this guitar part.” “The vocals in the second chorus are amazing.” “I think he screwed up the bass part there.”

Every kid was a music critic. Quality mattered. The music was not necessarily better quality, but engagement with music focused on quality.

Some of my Gen Z friends have told me they would consider that situation horrifying. “Why would I want to sit around and be forced to listen to other people’s bad music and having to make some kind of presentation about why I like my music? Why not just let everyone listen to the music they enjoy?”

Yeah. I get that.

Being immersed in music in your headphones is pretty awesome too. Those music sharing sessions got annoying sometimes. You had to sit though some bad music occasionally, and there was always that one jerk who would jump up saying “We have to hear that guitar solo one more time!”

I was that jerk on many occasions.

So, my Millennial and Gen Z readers, I’m absolutely not saying this was a better way to listen to music. I might get nostalgic about it, but I don’t listen to music this way any more. (Although, perhaps I should.) However, these constraints, as annoying as they could be, forced previous generations to engage with music very differently.

They forced listeners to consider and commit and engage with music in terms of quality. This is why every Boomer, Gen X kid, and all but the youngest Millennials will longingly tell you about the first album they ever bought with their own money. (Kansas, Point of Know Return)

When things got so hard for artists

Things began to change in the early 2000’s. Many blame the streaming services and the super-abundance of music they brought, but the change is deeper and more pervasive.

About that time, structure and affect started to get locked down across most mainstream culture from music to books to movies. This lock-in has nothing to do with AI, but it has set the stage for AI dominance of artistic output.

If you listen to the Spotify top ten, perhaps a third of the songs would not have sounded out of place if they hsd been released in the 80s or 90s.

This is unprecedented. The hits from the 80s sound nothing like the hits of the 40s. The hits from the 70s sound nothing like music from the 30s. Since the beginning of recorded music, every decade of popular music had a unique feel. A unique vibe. A unique affect.

Until now.

As an aside, I feel kind of cheated here. When I was a kid growing up in the 80s I looked forward to listening popular music 30 or 40 years in the future that I would hate just as much as my dad hated the Dead Kennedys. And that was a lot!

Today’s popular music does not lack quality. Taylor Swift is fine. Billie Eilish is amazing. But my dad would have liked Taylor Swift and Billie Eilish. I miss being challenged by popular music.

After a solid 100-year run, challenging listeners with new music is no longer profitable. Most listeners seem perfectly fine with that.

Spotify has been rightly criticised for steering listeners away from human artists and towards AI-generated songs. However, the business and engagement models of streaming services make this inevitable.

It’s going to get worse.

Spotify also illustrates how engagement with music has changed from one focused on quality to one focused on structure and affect.

In fact, Spotify is no longer about music at all. Spotify is about playlists.

Playlists are the primary means by which users interact with the platform, advertisers reach users, and users interact with each other. I would wager that many popular playlist curators haven’t even listened to all of the songs on their playlists.

Why should they? Who has that kind of time?

Listening to a song takes two or three minutes. In that time Spotify can have a user engage with 10 different playlists and three different ads. A user can put together and share a new playlist in less time than it takes to listen to two songs.

The reason so much writing, podcasting, and music is vulnerable to AI disruption is because quality has already become secondary.

The shift towards structure extends well beyond music. Most nonfiction books today have the structure and affect of a book, but they are not primarily intended to be read. Most are either a way of obtaining the media exposure a book-tour can bring, or as a way to establish the author as an authority on a topic.

Books, like music, have mostly become content.

Television news has the structure and affect of trusted news from 50 years ago, but discussions of journalistic quality are restricted to occasional hand-wringing segments, while the primary mission is creating news content. Unsurprisingly, AI is already widely used in producing financial and sports news.

Social media copies the structure and affect of community — people asking questions, sharing information and ideas — but ignores the quality of those interactions to focus on engagement metrics. Naturally, social media is already starting to be overwhelmed by AI.

The reason so much writing, podcasting, and music is vulnerable to AI disruption is because quality has already become secondary.

Structure and affect are the primary lens through which we view most creative works, and AI is incredibly good at copying the structure and affect of almost any kind of artistic work. That’s why the culture industry is so excited about it.

A Game Plan for Creatives

So, as an artist, how do you earn a living in an AI-dominated, content-focused world?

The following three point plan should put you on the right track. It won’t be easy, but making a living from your art never has been.

Granted, I might not be the ideal person to be give advice on becoming a successful artist. My music careers in both Tokyo and LA were not exactly chart-topping successes. I was Japan’s first professional podcaster in 2016, but I dropped the commercial side when I saw how the industry was changing.

However, I do know how to bring new technology to market. I’ve evaluated hundreds of business models and been deeply involved with dozens of product launches and technology go-to-market plans, so I can tell you exactly what AI is going to try to take away from you and what it won’t be able to.

This is the way forward

1. Never produce content

Learn to recognize when people are pushed to engage with structure and affect rather than quality, and don’t let feedback from those places guide artistic decisions.

Anytime the word “content” is used to describe artistic expression, it’s strong signal that structure and affect are primary. If your art is supported by advertising sold by others, then you are creating content — or worse, training data. Quality is irrelevant, and you will eventually find yourself in a losing battle against AI.

There are, of course, many quality podcasts and YouTube channels, but quality is irrelevant to the platforms themselves. Their goal is to create content with the proper structure and affect to sell as many ads or subscriptions as possible. If quality can do that, that’s fine. If something else can do it cheaper, that’s even better.

Once growth begins to slow, the profits of all ad-supported platforms depend on reducing the cost of the content. Creative platforms position themselves as communities of supportive creators, but they are more forthcoming in what they tell their advertisers.

You are a means to an end, and you will be cut out of the loop as soon as a cheaper means to that end can be identified. Right now, generative AI looks like those means.

Another tell that the interest is in structure and affect rather than quality is when conversations focus on fitness for purpose.

People looking for waiting-room background music, amazon product descriptions, or header images for social media posts are not engaging in terms of quality. The right structure and affect it is good enough. Anything more is a waste of money. That’s why AI is already dominating these areas.

You see this on Spotify as well. Many playlists focus on things like “Music for concentration”, or “Jazz for relaxation”, or “Classical for better sleep.” Listeners don’t want to engage in terms of quality. They just want to listen to a few minutes of something with the structure and affect of relaxing Jazz. AI-generated music is perfect for this.

The Jazz lovers are the only ones who want to talk about quality, and they are not part of the transaction.

The great commercial art of the past was created not because artists were better, but because technology was worse.

These motivations have always been there. All the buyers of commercial art have ever wanted was the structure and affect. They have always just wanted content.

In the past, however, the only way to get that content was to hire artists to make it, and a lot of those artists brought their A-game, not because they got paid more, but because that’s just what artists do.

That is why we have those amazing travel posters from the 20s and 30s, Schoolhouse Rock from the 70s, and Toulouse-Lautrec’s Moulin Rouge posters from the 1890s.

Neither buyers nor consumers demanded that level of quality. The institutions paying for it were not trying to create art that would endure for generations. They just wanted content, but the world got inspiring and enduring art as part of the bargain.

The great commercial art of the past was created not because artists were better, but because technology was worse.

Generative AI lets the buyers get their content without you.

2. Connect with fans on your terms

For musicians, an obvious way to connect directly with fans is live performance. Musicians have been making money from live performance for over 4,000 years, so it’s a well-proven method. It’s hard right now. Fewer venues support live music than in the past, and fewer still pay for it. Even the pros are feeling the pain as record labels are keeping more and more of the tour revenues for themselves.

Live performance, however, is worth cultivating, even if it will need to take a different commercial form in the future. The global phenomenon of Taylor Swift’s Eras tour shows that we still want to and need to connect with live music just as we have for thousand of years.

It’s a deep part of who we are.

Connecting online is important but complicated. Some creative platforms are powerful tools and you need to use them, but stay skeptical of promises to help you connect with your fans. Most of these platforms will start squeezing the artists hard once they have generated enough content for them.

They are not bad people. Their business model simply demands it.

When in doubt, look at what they say to their advertisers. If the focus is on content and structure and affect, then at least you know what you are signing up for.

Of course getting into a platform like YouTube, TikTok, or Spotify early, before the structure and affect gets locked down, can be fantastic creative opportunity. The early days of YouTube, Etsy, podcasting, and the internet in general was a glorious time of creative experimentation. No one knew what the structure and affect was supposed to be, so the only way to engage with what was being created was though quality.

Once structure and affect starts to get locked in, things change.

YouTube, TikTok and similar platforms can be part of your plan, but don’t count on them to make money. More important, absolutely never take their feedback on how to create better art. They will just lock you more tightly into their preferred structure and affect.

There are also a handful of companies that have structured their business models so users are incentivized to engage with creative works in terms of quality. Substack, Patreon, and Nebula fall into this category. Their business models incentivize connection to specific creators rather than casual content consumption, so at least their interests are aligned with yours.

That’s a good place to start.

3. Subvert structure and affect

Learn how to spot where structure and affect are becoming locked in and subvert it!

You have the edge here. LLMs cannot meaningfully subvert structure and affect. It goes against the very nature of what they are.

More important, society needs you to start subverting structure and affect. It’s the only way we are going the break out of the cultural slump we seem to be stuck in.

Subversion of structure and affect is actually an age-old creative technique. Satire is based on this subversion. Whether it is the Daily Show adopting and subverting the structure and affect of a news show to comment on our relationship to the the news or Blazing Saddles subverting the Western to comment on racism.

This subversion happens in music as well. John Cage wrote pieces like 4′33″ and As Slow as Possible that subverted both structure and affect to such a degree that listers are forced to ask themselves “Is this really music?” Punk rock kept the structure and subverted to affect to such a degree that my father’s generation emphatically insisted that “No! This is not music.”

That’s progress. That’s art.

Subverting structure or affect forces people to either ignore you or to engage with you in terms of quality.

It’s always simpler to give people what they expect and what they think they want, but AI will soon be able to do that better than you can.

Seek out places where structure and affect has started to settle in and subvert it by doing something different.

And to the musicians out there. Please create something your peers consider visionary and that old-timers like me will absolutely hate.

Our cultural future depends on it.

A request to listeners and readers

I also have a favor to ask of music, book, podcast, and movie lovers.

Just as the whole world befits from the creative output of a small number of artists, it also benefits from a relatively small number of enthusiasts who may not be artists themselves, but who know what they like and can explain why they like it.

Most people don’t take the time to consciously look beyond structure and affect and deeply engage with quality, but subconsciously most of us can tell the difference.

That is why quality art endures for decades or centuries — sometimes being rediscovered years after the creator’s death. It is also why low-quality creative works, even those that are global sensations, quickly fade from the culture.

The enthusiasts’ ongoing engagement with quality makes that happen.

The favor I ask of you is to become a critic. Don’t adopt the structure and affect of a critic. There is no need to wear a tweed jacket and glasses or smoke a pipe. Just start engaging with quality. Know what you like and why, and tell people about it in real life.

You don’t have to start listening to opera or jazz or visiting art museums — although there are some pretty cool things to be discovered there. Keep watching, reading, and listening to what you do now, but engage with the quality.

It’s harder than it sounds. Engaging with a creative work in terms of structure and affect can be done in seconds. Engaging in terms of quality takes time. You’ll be forcing yourself to consume fewer creative works, and that can be challenging in our binge-watching culture.

Carve out some time to just listen to the music you love. When you watch a movie let the CGI and witty banter just wash over you, and see if the human story resonates with you. Discover what moves you, and why. There are no wrong answers.

You will discover two things pretty quickly.

First, you’ll find the second screen is your enemy. Doomscrolling while listening to music or watching a movie will cause you to disengage and overlook a lot of the quality that can make a work great. Despite what we like to tell ourselves, science has repeatedly proven that we can’t actually multitask this way.

Second, you’ll discover that most of books, movies, and music are not worth spending that kind of time on. Turn them off and keep looking. When you find something that moves you, share it with the people around you in real-life (not just on the internet) and let them know why you like it. Ask them what they like.

This is how discovery actually works.

One of the biggest disappointments in music streaming has been their promises of discovery. The claim was that with millions of tracks from thousands of artists with uncountable numbers of global influences, we would discover all kinds of new and exciting music.

But that never happened.

From junior high school onward, I have discovered far more new music by talking with people in real life than I have on platforms like Spotify or iTunes. Spotify feeds me playlist after playlist of new tracks and new artists, but it’s the same music. The same structure. The same affect. Quality is irrelevant.

I believe the problem stems from the fact that when streaming gave us access to musicians from all over the world, it also gave those musicians access to a global audience. When you combine that with an algorithm that instantly rewards and punishes based on engagement, and a arms race of optimizing structure and affect naturally result.

Perhaps innovative new styles need to develop in isolation; emerging and being refined among like-minded visionaries and supported by a handful of fans and enthusiasts who really get it.

Those handful of enthusiasts engaging with artistic works in terms of quality are the ones who not only bring innovative new works to the attention of the rest of us, but but just as important, they determine which works endure.

Epilogue: Can AI create “real” art?

There has been a lot of discussion about whether AI can create genuine art.

LLMs will not. They are structure an affect imitation machines. That’s just how they are architected.

However, long before we reach anything close to AGI, AI will evolve to the point where it develops intention; where it can travel the net, interact with the world, and build up its own private experiences. With so much information comprising their rudimentary consciousness, I can absolutely see AI developing a desire or compulsion to create art.

We might not be able to understand such art. It would most likely have a structure and affect that we won’t even recognize at first.

But the AIs would not be producing it for us. They would be producing it for each other, or perhaps, just for themselves.

Resources

Here are some people using various media whose thinking on this topic is worth engaging with deeply.

• Ted Gioia’s substack The Honest Broker is a treasure trove of insights on the state of popular culture and the history of music.
• Rick Beato’s YouTube channel often explores the state of modern music and the reasons behind it.
• The Philosophize This podcast had a great episode on the Frankfurt School’s thinking on how technology changes our interactions with music and movies. Part 1: Part 2
• Max Horkheimer’s Dialectic of Enlightenment introduced the concept of the culture industry

And of course, check out Disrupting Japan, my own podcast on startups and innovation in Japan.