What does it mean to be a musician in the age of AI?
The last of my 'music' series... for now!
Things of Internet is a newsletter operating at the intersection of marketing, technology, culture and humanity. I’m Deepak aka Chuck, and I straddle marketing, content creation, creator coching, a varied content diet, and overthinking of all the above.
A month back, I wrote about what it means to be a musician in the internet era (Part 1, Part 2). In case you need a recap of my snazzy art-in-Powerpoint skills, here’s that spectrum thingy I made up.
I went down this road as I started wondering what it means to be a musician today - not so much the internet age, but the AI age. So I dove into the past to try and get some lessons to extrapolate to tomorrow. Along the way, I ended up reading a lot of nice articles, watching many YouTube videos, and (most scientifically), devouring several Instagram comments - the underrated sentiment source for The Modern Researcher.
One observation I have is that when AI proliferates coding, marketing and tech, the reaction is largely lamentation over job losses or rants about capitalism. But when AI infiltrates art - oh boy! Reactions tend to be a lot more visceral, and not just by artists. Even reactions to Coke’s AI-generated Christmas ad (closest to “art” in Big Commerce, I guess) were less about cost-cutting and more about loss of soul.
And I think that’s a good place for us to begin this piece. Art and artistry means something more to us, and will be preserved, no matter what. The German art historian Erwin Panofsky has a nice framework to think about this. In short, it says that any art is interpreted / evaluated by humans at three levels:
The primary meaning - surface-level. Like the literal landscape in a painting, or a chord pattern in music.
The secondary meaning - symbolic or cultural meaning. Like what the painting denotes, or abiding by musical genre conventions.
The tertiary meaning - this is something more intrinsic. The emotional meaning, the back story, the human experience that led to the piece of art.
A simplistic way of thinking about AI art, then, is that it lacks tertiary meaning, even if it excels in the other two. Turns out, this matters a whole lot to us humans. The art we truly like, we do not because it sounds good or uses good brushstroke technique (that helps) - but because of the intention behind it. So we can say that no matter how good LLMs get and how many NVidia processors Suno buys, purely machine-made music will always lack that all-so-important tertiary meaning.
Or to abstract that to simple terms: Humans appreciate art made by other humans, even if it is ‘imperfect’.
I can vibe with that. Humans appreciate effort over polish - at least for art1. For example, I am hoping that you (human) appreciate the effort I (human) took for that silly illustration above despite the lack of polish and the almost-typo in the last line.
Functional vs Emotional art
When technological advancements / existential crises come for art, it often forces fundamental questions like… what is art or who even is an artist, as I’ve tried to explore so far in this series.
I think the answer has two facets - functional and emotional (TBH it lies on a spectrum, but enough of that).
Functional art - stock music, perfunctory landscapes in 3-star hotels, representative visuals in many publications - is at risk. Not just from AI, but other technologies. If we’re using the “music to study to” playlist on Spotify today, swapping this with a headset that magically fires up your synapses to retain information is a logical choice. Homo economicus makes this choice. Technically, ‘music’ is dying here but I doubt too many tears will be shed over this.
Those who choose to, however, listen to only Iron Maiden while working out because they and the emotions they conjure up (rather than the high beats-per-minute, energetic scales etc) get you totally fired up for another round of burpees - would not be caught dead with such a gadget. Which brings us nicely to the emotional end of the spectrum. This is something I feel we will always crave and even seek out. AI or not. (And maybe moreso in the age of AI).
All of us care about some art enough to proactively seek it out.
Case in point - I am part of (actually, founder of) a thriving music discussion group on Whatsapp where people share hundreds of recommendations weekly. I’ve found new music from artists I already love, and find new music there all the time. It’s fabulous for random-yet-curated discovery. And yes, I’d rather that over a “more likely you’ll find something you like based on what we know of you” Spotify algo. It may not be perfect, but it’s part of the “what it means to be a true music fan” life. Seeing other people’s fandoms, getting their recommendations, writing back to them, forming a community, and all that.
Similarly, you may care enough about, say, sci-fi novels, renaissance-era paintings, or architecture enough to proactively seek out sources, community and nerd out. Yeah, you MAY come across the perfect song / fiction book / building but it will still lack the all-so-important tertiary meaning that will turn it from perfect to perfect for you.
For functional art, the tertiary meaning does not really matter. Or in other words, graphic designers who make social media creatives for a living are at risk; while those who are commissioned for portraits for a known style, are not (this I know is a very debatable matter but I hope you see my larger point, please don’t quote me out of context, I have many graphic designer friends).
Enough rambling Chuck. Tell me 5 things that will happen to music / musicians in the age of AI.
Yeah okay. Here goes.
1. AI will be used in music more - duh
In the recent Pink Floyd Pompeii film2, there’s a very telling scene where guitarist David Gilmour spoke about all the fancy new technology the band was using (early 1970s!). He said something to the effect of:
Yeah, the tech is all there, but the idea, the music, still needs to be formed in your head. All the tech is just a way of getting it out, isn’t it?
I feel putting that very question in was an editing choice given it came from what is considered today to be an OG band, in the age of AI. I am imagining many people who have ideas for music in their heads - be it simple melodies or whole orchestral suites - but no musical playing ‘skill’ will be able to put it out. And hey if it moves people, why not?3
The parallel in technology we’re seeing is vibe-coding. Some people hate it, I’m all for it. Is it perfect code? No. Is it a great way to showcase an idea and prototype it? Heck yes. I actually have a few ideas that I hope to vibe-code and unleash on you folks some day.
The reaction to vibe-composed music will be interesting. Already, many people hate the idea of generative AI art for a combination of ethical and aesthetic reasons. But today’s music already uses so much AI and other technology. We’ve come to accept sampling - at its core way more egregious than Generative AI.
I, as a neutral observer, look forward to seeing what kind of creativity can come out when creativity meets technology. As a friend (and the LAST person to pay me for a Things of Internet subscription in its paid era :D) told me:
Btw one aspect of Suno for me is that it’s really a 21st century musical instrument. It fascinates me that the next few years for it will be about moving beyond prompts and using batshit crazy smart phone interactions (touch, sound, visual, camera etc) to somehow extract the music in people’s minds and output it as a file.
One hundred percent agreed with that.
PS - this is NOT to diss people who actually play instruments or compose the old-fashioned way. Liking one does not equate to negating the other. In fact, I still hold that in an age of computer-made music, a live experience with real instruments becomes even more valuable.
2. “Real” musicians must lean into the human element even more.
What I mean by this is - live shows, interviews, behind the scenes… but also more emotion, more vulnerability, more personality, less plastic.
In fact, I would go to the extent of saying that a lot of highly manufactured pop today will be killed by generative AI - the corporations that give us soul-less music with manufactured artists may as well do that with virtual artists. The rest of us will cling on to our actual musicians, thank you very much.
But back to the point. For a musician to survive today, they need to lean in to their humanity - be it through their music or their social media skills.
A weird example I have is the support that Axl Rose got for slipping on stage at the recent Guns n Roses Mumbai concert - many in comments talking about how it showed this was a real moment, an unrehearsed thing, something you don’t see in highly manufactured productions, and so on.
3. Generative AI will help with some part of the non-songwriting process
This could be everything from musical (mixing, mastering) to the ancillary (creating social media content out of a whole video).
And why not? There are tools that replicate an entire recording studio, that have enabled a generation of bedroom music makers to flourish (as always, some good and some bad results).
I use GoPro’s AI tool to find good clips from a 40 minute dive video and it does a pretty decent job.
So even if you baulk at AI working its way into the actual music (or are afraid your fans will cancel you for daring to use Suno), there is still a way AI can help your workflow.
4. We’ll see new musician roles
The movie era gave us film scorers. The gaming era gave us game music specialists.
And while the AI era will, yes, take away some functional music jobs, it also has the possibility to give us some new ones. I foresee an explosion in vibe-coded games and apps, each with their own nuance, many of whom will want more than just AI-generated music… Especially if it moves on from prototype stage, and it wants to cross the ‘tertiary’ border to appeal to actual humans. I’m bullish on this. If you’re a musician, please don’t stop playing your guitar (or, at the very least, thinking musically).
5. Originality will continue to matter.
As you’ve probably gauged from my tone, I am not anti-AI.
But at the same time, there is NO way that any any LLM will be able to come up with Bohemian Rhapsody. It might NOW, when the song is established canon… But not in 1975 when it was written and was something so fantastical and crazy that… well, the reaction of the EMI exec in the scene from the eponymous movie says it all.
Neither could AI have come up with Tomorrow Never Knows in 1966, where The Beatles went absolutely manic in the studio experimenting with all kinds of things to create something that sounded like nothing before it.
LLMs - image, text, audio, video - are trained on existing data. If your ‘data’ is something that never existed, it can’t cook it up. Unless you have the idea in your head and prompt it in imaginative ways.
If you’re a djent band sounding like the hundreds of other 8th string-chugging acts peddling the same hackneyed melody-meets-dark fare (and probably with a single plural word name and album art with triangles and stars), you don’t need AI to render you obsolete.
It’s instructive to look at “____ in the style of ____” AI cover videos. The top ‘styles' always seem to be artists who have a very original style. Linkin Park. System of a Down. Metallica. The Beatles. Queen. Nirvana. Maybe that’s something to think about. And heck, maybe ‘style’ will get copywrited in the future for use in AI models just like how lyrics and melodies are, in today’s paradigm.
Net-net - originality matters, AI or not. Becuase the ultimate audience is a human, not a machine.
There were many songs I liked in 2023. Metallica’s Inamorata became a top 5. Indian Ocean’s new Jaadu Maya was gold. Fall Out Boy’s Heartbreak Feels So Good was a pop masterpiece. But. The one song that defined the year for me was this one4
When I first saw the thumbnail, I was incredulous. 3 minutes later, I was almost in tears. It sounded perfect, Chester’s screams were on point, there were thoughtful additions that only a true fan could have put in… I found myself humming this in the Linkin Park style for the rest of the year… and upto now. Yeah yeah it was made using AI, but it fucking hit me in the feels. Isn’t that what art is supposed to do, no matter how it was made?
I don’t think it’s healthy to adopt a “it was made with AI so f*ck it” attitude. I think it deserves a bit more nuance than that. At the same time I’m not saying we should readily embrace everything an LLM spits out. These are tools - judge the output, judge the intent, and question fundamentals. What of that is something that only humans can do? What of that is something that’s just “the way it was done” and we’re being precious holding on to? What is an opportunity?
See, there is no doubt that music and all of art is going to be affected by AI. Some positive, some negative. We’ll see new roles and laws, society having new debates, and the emergence of new art forms. We’ll grapple with the future, as we ask some very fundamental questions (what is art, what does it mean to be human).
And wading into all that is what I as an observer look forward to. The intersection of music, technology and humanity is something that will be a recurring theme in this newsletter. And I hope you’ll be there for the ride. Thank you for reading this far!5
Other things from the internet
McKinsey has a newsletter on GenZ. What a time to be alive.
My nomination for ironic banner ad of the week:
In #thisexists: Did you know there’s a 4000+ strong subreddit called F*ckZepto? Anyway, here’s a gem from there:
See you next week! Listen to some human-made music this week :)
Chuck
A fundamental takeaway from Rory Sutherland’s excellent Alchemy.
A watch ONLY for hardcores - please don’t watch if you’re hoping for Coming Back To Life.
Yes, I get there are ethical issues, copyrights and all that but I believe those will get sorted out in time. I imagine in the future Suno will feature a combination of original music from ‘functional music makers’ and actually licensed music. The current debate we’re having is an important one that will lead to solutions like this, just like how the initial outrage around sampling eventually led to a transparent and ultimately effective process.
Unfortunately, the ‘original’ version of this, uploaded by a MrPooby, was taken down for, sigh, copyright violation. Thankfully I have an MP3 version downloaded to revisit it, and it still floats around on the internet.
While I’m concluding my music series here, I’m pretty sure I’ll revisit this topic again. I’m way too interested in it, and so much progress is happening in the space. I hope to do some musical experiments myself, and speak to others who are actually involved in the space. Maybe I’ve been too flippant, and I hope to be corrected by some musicians (all artists, really) who’ve actually been affected by AI. Reach out. Would love to talk.