AI Music, simply explained (feat. Grimes and Spotify's CEO)

When should artists get paid in a world with AI music?

Artificial intelligence is changing how music gets made - and how musicians get paid. AI is letting people clone artists’ voices, create completely new songs as fake collaborations, generate lyrics in seconds, even produce full tracks just by typing in a few words. It’s all causing some to say AI will be “the death of music.”

Technology causes turning points in history. And I think we’re in one right now for music. The stakes are high: If we get this wrong, we could jeopardize how human musicians make money and art. But if we get it right, we have an opportunity to leap ahead in how we as humans get to express ourselves.

To really understand what’s happening with AI music, you need to understand how the music industry ALREADY works - and how it could be changing. In this video, I took a deep dive into that topic with the help of two people right in the thick of it: The artist Grimes and CEO of Spotify Daniel Ek.

Chapters:
00:00 What is happening with AI in music?
02:05 Why would we want AI music?
04:03 How are musicians using AI?
06:17 When should musicians get paid?
07:15 What is copying versus inspiration?
08:26 How does copyright work in music?
10:06 Should artists get paid for AI training on their music?
11:55 What happens when AI generates its own songs?
12:54 How could AI music be "huge if true"?

Bio:
Cleo Abram is an Emmy-nominated independent video journalist. On her show, Huge If True, Cleo explores complex technology topics with rigor and optimism, helping her audience understand the world around them and see positive futures they can help build. Before going independent, Cleo was a video producer for Vox. She wrote and directed the Coding and Diamonds episodes of Vox’s Netflix show, Explained. She produced videos for Vox’s popular YouTube channel, was the host and senior producer of Vox’s first ever daily show, Answered, and was co-host and producer of Vox’s YouTube Originals show, Glad You Asked.

Additional reading and watching:

Gear I use:
Camera: Sony A7SIII
Lens: Sony 16–35 mm F2.8 GM and 35mm prime
Audio: Sennheiser SK AVX

I used music from Lickd for this video:

License ID: 81a29AAvBJO

License ID: 2ZnbrkkMnz1

License ID: nYJ7D662Gqk

I also used music from Tom Fox and Musicbed!

Special thanks to Daniel Ek and Grimes for their thoughts on this conversation!
Thank you also to Angela Long for inspiring some of the graphics in this episode, based on a talk I did on this topic recently!

—
Welcome to the joke down low:

How do you fix a broken tuba?
With a tuba glue…

Use the word “tuba” in a comment to tell me you’re a real one who read to the end… :)

[wishhhv]

As someone who's an audio machine learning engineer and a part-time touring musician, I don't share the same optimism:
1) Technically speaking, these algorithms don't "understand" or "compose" music, they simply capture different audio spectrogram patterns in the input audio data. I think it won't replace musicians and might even partly assist in compositions, but these algorithms learn over time. So, what's stopping Spotify from flooding itself with subpar music that's just training data.
2) Fundamentally, these arguments are like "go to an expensive restaurant and the chef just gives you a pill which makes you feel the same way amazing food does", or "pay expensive stadium game tickets and you only feel the excitement of a great game without actually watching the game". But basically these experiences are meaningless, just an illusion of the real human experience.
3) Making music should be at least a bit difficult, otherwise we start getting formulaic similar (arguably mediocre) music, which is what we're witnessing with some of the very popular names. This isn't some new instrument or a new recording technique or tool. This is LITERALLY algorithmic shuffling of different compositions to spit out a mindless "tune" that only sounds humane, but actually isn't. Make royalty distributions higher, fairer, and more transparent. The problem lies with these for-profit organizations that'll always prioritize profit over creativity & supporting musicians.
4) Music education standards are so poor that people just don't understand how difficult it is to make ANY kind of composition. And don't get me started on the frustrating system of music payments & royalties (Spotify is one of the WORST offenders!).
5) This is how dystopia begins, wherein we value a mindless algorithm that's devoid of human emotion to churn BS instead of valuing artists who've poured their lives into creating something new.

[emory]

I find it curious that you are using the CEO of spotify to talk about compensating artists properly with Ai stuff, when spotify themselves is one of the worst offenders for compensating artists in general.

[uasiddiq]

Cleo went all out with this video. Great job not letting the embarrassment get in the way 😊

[HristoVelev]

Spotify's endgame would be to have a recommendation type algo feed the AI and serve you music, and cut the artists out of the picture. For the human expression part they would hire some actors or use generative AI for videos - all payrolled by Spotify, and all rights belonging to them.

Great video!

[vsnrm]

Interesting how artists were called overdramatic for reacting to the AI apocalypse but when it's affecting other fields of art suddenly everyone takes it seriously. Just find that interesting

[Telleryn]

The biggest problem I see with AI in any creative space is not that it can make things, but the amount of that thing it can create, an AI can create thousands of songs in the time it takes a human to write and record one and out-compete humans just through sheer volume, burying human work under piles of AI stuff and making it undiscoverable, even things like tags etc that might be used to highlight human-made work only work as long as those uploading AI-produced work are being honest.
With how easily recorded works can be produced by AI now, we might see a return to how musicians got their fame and money before recording was widely available, performance, before AI comes for that too.

[scott-richardson]

What’s more profitable? Paying artists royalties? Or having an AI generate millions of new songs for your platform that you own? Suddenly you’re paying royalties to yourself, Spotify!

[estnis]

Pretty sure everyone agrees do NOT get fairly compensated. Yeah, not even Prince, Michael Jackson or Taylor Swift (to give three well know examples of record label disputes). So what makes us think that if we throw a very complex and non human variable to this equation, all of the sudden all these humans will be paid fairly? Not gonna happen.

[jaredkhan]

I’m a musician, and I definitely do fear this topic. AI can be an incredible tool for filling in the gaps in like mixing and mastering (like LANDR), but what makes a musician is not only creativity but skill as well. Dedicating hours to training ur body to play an instrument, sing, learn your DAW, learning composition, etc. are all part of the journey of being a musician.

I already kind of have this stance with sampling. I hear some amazing and creative uses of samples (I use them myself, as does practically everyone), but the average “beat maker” truly feels like a talentless hack. I’ve watched ppl just put a beautiful sounding loop in with some basic 808 pattern and call themselves talented with no understanding of why that sample sounds good to begin with. Letting AI make lyrics, make ur melodies, etc., it all just seems like a mask for ppl with no skill or dedication to pretend to be skilled.

Ik I m probably sound elitist or gatekeeping but it’s just my perspective given my experience.

[joshualane]

Having a discussion about the artists getting paid, featuring the CEO of Spotify, but not talking about how much artists are actually getting paid for their work is wild.

[gautambidari]

Props to the editor of the episode. It's absolutely beautiful, such a good composition.

[ericwood]

I love how you ask the right questions. Great journalism!!

[Dominorican_naturalist]

What makes art amazing to me is the amount of time it takes an artist to perfect a skill, the human error, the whole process. It looses value if something generates perfection every time in my opinion. Being human is fun and full of mistakes and lack of perfect.. AI for creativity is a buzzkill for those who have trained for years just so people who haven’t trained a skill can seem just as competent in that area.

[ChristianSantiagophoto]

I think subtracting the skill it takes to make art dilutes it. The process of making art, the suffering and effort behind it, the journey to get there, adds so much to the final outcome. If any dingus can make beautiful music or jaw dropping images, then they cease to be special.

[bentownsend]

All previously invented technologies have been tools for artists to use. AI is the first tool to be able to replace the artist themself.

[Krustenkaese]

to be totally honest the last people I want to hear from when it comes to this are both the Spotify CEO and Grimes.
There is no way that guy isn't going to use AI to make more money for himself without giving the artists their fair share (as he has been doing for years). Plus, when it comes to Grimes, someone so technology obsessed who only sees the benefits but isn't interested in considering what it'll do to artists who aren't mega rich, I'm just not interested to hear what her enlightened take is.

[Glouryian]

Somewhere I read something like "People working for minimum wage while machines make poetry and art is not the future I imagined."
I would like to feel different about it, but I deeply worry that the enthusiastic exitement around AI is naiv or even worse harmful. Enabling people to realize whats in their head sounds like a huge selling point. But AI does not create any knowledge, but dependence. Someone who does not have the abilities to write a song will still not be able to write a song but be dependent on AI tools to create something.
Education and training are the tools which really enable and make people independent to express their own unique creativity. The struggle to make an idea into something others can enjoy is part of the creative process, because it forces you to think differently, to find solutions. Each work of art is the result of the whole unique human experience of the artist, a series of thousands of individual experiences and creative choices. Generating AI skips all of that. AI may create you a song, but not the song, YOU would have created. And thats a huge difference.
And for the copyright part: the human inspiration and the AI "inspiration" is not the same. If you paint something in the style of van Gogh the outcome will still be highly influenced by your personal experience of the art. Maybe you will focus more on the paintstrokes, or the colours, or the light. Maybe you connect van Gogh to a certain emotion or memory. In the end you will make a lot of individual creative decisions which will lead to your own unique work of art. AI does not does not have any that.
The worst outcome could be that in the future the creative monopol may lie in the hands of AI coorporations. In the capitalist world we live the human factor is the worst for business. Artists can have a creative crisis, may get sick, may sometimes be not easy to deal with. Now there is a solution which steadily creates vast amounts of output without any of these problems and of course this has the potential to push aside human creators on the long term. And creators may become dependent on their tools to realize their artistic ideas.
There of course will still be musicians, artists, writers who will be stars and create unique masterpieces, but for vast majority of small, independent and unique thinking artists this will make a situation much harder, which is already much less than favourable.

[sayGIRLIMNOTFALLINGFORIT]

I prefer raw music, raw human emotion.

[skycladsquirrel]

I have been preaching this tech for over 20 years. The business model needs to make sure that artists and writers get payments for the songs we create. Then everyone will be happy. Imagine getting money for something you never made. Plus, If an artist likes the collabs could sing them in concerts and make even more money. It's exciting. But we have to make sure it's done right.

[mkst]

That animation showing how an AI generator could come up with a million tracks and then plot a path to ‘success’ was chilling. The amount of dreck that’s going to be produced is going to be mindblowing.