Can AI sokkie?

What AI music might mean for the (Afrikaans) music industry

He’s seen your braai skills; now show him your reading list. Give your father-in-law an Our Long Walk gift subscription this Christmas, and upgrade every Sunday lunch conversation. Set it up in seconds.

I composed a song in Afrikaans. (Now available on Spotify.) The thing is, its pretty good. And, of course, I did it with AI.

Over the last few weeks, AI-generated Afrikaans music has been making it up the charts, finding airtime on radio and listens on platforms like Spotify. It’s also not that surprising: globally, AI music is increasingly popular – and lucrative. AI R&B artist Xania Monet (listen here) reportedly signed a $3 million deal recently. Several musicians responded with, well, disgust.

Afrikaans artists, composers and producers are worried that AI-frikaans music will displace their jobs. But I think this view is too pessimistic. In fact, I think it is exactly the opposite of what will happen.

Let me stick my neck out: there has never been a better time to be an Afrikaans musician if you are willing to work with AI.

Think back to the 1970s, when automated teller machines appeared on the street corner. Many predicted the end of bank tellers. Machines could now do what humans did: hand out cash, take deposits, handle everyday transactions. Yet that is not what happened. As the economist James Bessen has shown, the number of bank tellers in the United States actually rose in the decades after ATMs arrived.

Why? ATMs made it cheaper to open branches, so banks opened more of them. The machines took over the boring, mechanical work. Tellers spent more time on the human work: talking to clients, giving advice, selling products. Automation did not wipe them out; it changed their job and, in the process, helped the industry grow.

Now imagine we stop thinking of ‘musician’ as one job. Some artists are loved for their lyrics. Others for the energy they bring to a stage. At the moment we expect someone like Karen Zoid to do everything: write, compose, arrange, mix, master and then still carry a room on a Friday night. AI can cut the cost of almost everything except the scarce bit she does best. She can focus on performance – the thing only she can do – while drawing on a huge catalogue of lyrics and melodies that she selects, tweaks and makes her own.

Once writing becomes cheap, the amount of material explodes. When that happens, the real scarcity shifts to the things machines cannot fake: presence, voice, timing, connection. That is what opens the door to mass customisation. You want a score for a documentary about your best friend Wally the Labrador? Easy. A wedding where every song is about the couple, written to echo their story? Also possible.

As composition becomes abundant, bespoke performance becomes where the value sits. You want a song for your daughter’s 18th, performed live by Karen? That will be expensive, but it is no longer a fantasy. In economist-speak, the ‘rents’ move towards live, embodied experiences that cannot be copied at zero cost. As Afrikaans musician Bouwer Bosch put it recently on Instagram: ‘die koring sal van die kaf geskei word as dit kom by toer’. Touring is one filter. So are small venues, private shows, personalised sets. Some of those experiences will start to look like classic status goods: the more exclusive they are, the more people are willing to pay to signal taste and standing.

So expect the payoffs in music to shift. The extra reward for writing one more generic lyric will fall as the supply grows. The reward for interpretation and performance will rise. Artists with real stage personality, a recognisable timbre and the ability to learn new material quickly will do well. The machines widen the pipeline. The human on stage remains the bottleneck. And that is exactly where the money tends to go.

There is another big upside to AI that we do not talk about enough: better quality. There will be AI slop – sound pollution – but it is not as if we have never had slop before. (Exhibit A). My claim is that when far more people can experiment, the overall quality of lyrics and compositions rises. Open the gates and you do not just get more noise; you get more draws in the creative lottery. When the cost of trying falls, the spread of outcomes widens, and with enough attempts the ‘upper tail’ thickens. That is just the economics of search under uncertainty: cut the price of exploration and you shift more probability toward rare hits. Some of those hits will be algorithmic mash-ups. The best will be genuine discoveries from people who would never have seen the inside of a studio in the old system.

Quality improves not because machines turn into poets, but because they act as force multipliers for human taste. A songwriter who might have managed three sketches in a month can now generate and sift through three hundred. They can then spend their scarce time polishing the two that feel like ‘hits’. This is logic of evolution applied to culture: if survival of the fittest song depends on a large, cheap pool of contenders plus ruthless culling, AI gives you both.

Platforms sit on top of this. Recommendation systems get a bad name, but at heart they are continuous taste experiments. They test which songs people skip, which they replay, which lyric makes them stop. If creators can see detailed feedback, they can iterate faster. There is a real risk of everything regressing to the bland average. The counterweight is audience diversity. Niche demand used to die at the radio playlist. With AI cutting both production and search costs, those niches can become sustainable micro-markets. That is also a kind of quality improvement: better matching between songs and listeners, fewer talented misfits left unheard.

Of course, none of this guarantees a musical golden age. Cheap production can tempt labels into high-volume strategies that drown discovery. Legal fights over training data can slow things down. Spam can clog the pipes. But lower costs move us from a world where production is the main constraint to a world where selection is the constraint. In the old regime you improved quality by pouring money into the few projects you could afford. In the new regime you improve quality by widening the funnel and upgrading the selectors – the people who can tell signal from noise and are willing to back unusual work.

For South African music, and Afrikaans music in particular, the upside is hard to miss. We sit on a deep reservoir of stories, idioms and rhythms that no global model will ever quite get right without local hands on the wheel. Put cheap, powerful tools in the hands of thousands of young creators. Match them with producers who are willing to cut ruthlessly. Get the best tracks onto stages where presence and performance still matter.

There will still be slop, plenty of it. But there will also be more outliers, more strange and interesting blends, more new stars. AI pushes down the cost of trying. Human judgement and live performance push up the rewards for the few who get it right.

Back to Wanneer, my sokkie treffer. What is it about? Well, AI and its impact on us.

I was inspired by a new paper – Machine Intelligence and Human Judgment – written by three prominent economists in the AI space: Joshua Gans, Ajay Agrawal, and Avi Goldfarb. I asked ChatGPT5 (pro version) the following, and attached the paper Gans et al.

I want you to write lyrics for a song. It must be in Afrikaans. The lyrics must combine the wisdom of Rudyard Kipling’s ‘If’ poem, but it must be applied to AI. In Afrikaans, the title must be ‘Wanneer’. Be inspired by Koos du Plessis and Koos Kombuis I attach a paper that should provide depth to the lyrics. Could you write the lyrics - three verses interspersed with a beautiful, memorable chorus, and a bridge as well?

It produced beautiful lyrics, though I had to tweak it here and there. Here is the English translation:

When the machine whispers to you: I know, and the world’s voices sink to a dull hiss, you must find the silence where your own judgement lives. For what is information without the hand that weighs its worth, without the eye that discerns the light from the dark gleam of a screen?

They speak of an island full of geniuses, night and day, at work, like slave-ships of light, and we carry their cargo across the oceans of money. But who holds the helm when the storms of inequality come? Who still knows the way when every map is drawn by algorithms?

In the towns where only a dim light burns the metal mind can be like bread, part of a new day. But in the towers of power judgement is honed like a sword. Whoever passes judgement owns the world as well; and whoever shares it can still plant hope in dust.

Bridge:

Child – you must remember: the machine sings without a mouth, but you carry the breath, you carry the blood. If you can listen, yet still choose, then you hand the future back to human hands.

Chorus:

When you walk with the machine’s shadow without selling your own; when you know: prediction is cheap, but judgement is dear, and only the human pays the price – then you remain free, then you remain human; then you remain free, then you remain human.

I then fed Suno the (Afrikaans) lyrics, and asked it to compose a song inspired by the following styles: countrypolitan, baroque folk, dream pop, American primitive guitar, accordion, ukelele, d drop, soulful, scratchy male voice, choir.

And boom, within two songs, I found my treffer. My artist name is Koos Komper. And the song is Wanneer. (I’ve uploaded to Spotify and other music platforms, though it is taking much longer to list there than expected.)

Can AI sokkie? In the studio the music will be abundant, on the stage it will stay scarce – and that is where we remain human. I will see you on the dance floor, AI music.

Wanneer (Koos Komper)

Wanneer die masjien jou fluister: ek weet,

en die wêreld se stemme word vaal geraas,

moet jy die stilte vind

waarin jou eie oordeel leef.

Want wat is inligting sonder die hand

wat die waarde weeg,

sonder die oog wat die lig onderskei

van die donker glans van ’n skerm?

Wanneer jy met die masjien se skadu loop

sonder om jou eie skadu prys te gee;

Wanneer jy weet:

voorspelling is goedkoop,

maar oordeel is duur,

en net die mens betaal die prys—

dan bly jy vry,

dan bly jy mens.

dan bly jy vry,

dan bly jy mens.

Hulle praat van ’n eiland vol genieë,

nag en dag, aan die werk,

soos slaweskepe van lig,

en ons dra hulle vrag

oor die oseane van geld

Maar wie hou die roer vas, wanneer

die storme van ongelykheid kom?

Wie ken nog die pad

as elke kaart deur algoritmes geteken word?

Wanneer jy met die masjien se skadu loop

sonder om jou eie skadu prys te gee;

Wanneer jy weet:

voorspelling is goedkoop,

maar oordeel is duur,

en net die mens betaal die prys—

dan bly jy vry,

dan bly jy mens.

dan bly jy vry,

dan bly jy mens.

In die dorpe waar net ’n dowwe lig brand

kan die blikbrein soos brood wees,

deel van ’n nuwe dag.

Maar in die torings van mag

oordeel geslyp word soos ’n swaard.

Wie die ordeel vel,

besit ook die wêreld;

en wie dit deel,

kan nog hoop plant in stof.

Kind - jy moet onthou:

die masjien sing sonder mond,

maar jy dra die asem,

jy dra die bloed.

As jy kan luister,

maar steeds kies,

dan gee jy die toekoms terug

aan mensehande.

Wanneer jy met die masjien se skadu loop

sonder om jou eie skadu prys te gee;

Wanneer jy weet:

voorspelling is goedkoop,

maar oordeel is duur,

en net die mens betaal die prys—

dan bly jy vry,

dan bly jy mens.

dan bly jy mens.

‘Can AI sokkie?’ was first published on Our Long Walk. The images were created with Midjourney v7.