The Year AI Broke Into Music

Bloomberg Audio Studios, podcasts, radio news.

Nearly twenty years ago, Grammy Award winning producer Timbaland got in the studio and made a song that would go on to become one of his most iconic from clubs and wedding receptions to Bob mitzvahs and everything in between. If you were alive in two thousand and seven, you couldn't escape its hypnotic.

Beat, unique.

De'spadiation right that song, The Way I Are, would peak at number three on the Billboard Hot one hundred chart. Critics were into it. That memorable synth line felt like a peek into the future. When I talked to Timbaland earlier this month, he told me he's always tried to embrace new technology in his music production. For The Way I Are, he and his team experimented with mix effects and synthesizers.

I just felt and I'm like, ooh, that's a dope pattern. So it's steal my taste.

Timberland made that song in his thirties. He's fifty two. Today. A lot's changed in music production. Sense. New tools have introduced new questions about how the over twenty eight billion dollar reported music industry makes its core product and artists, whether they're making songs in their bedrooms or at a big time music studio, have had to grapple with those shifts. Today, Timberland is an outspoken advocate for the tech that he thinks is going to most change the music industry next AI.

I look at it as an amazing too that I wish it was available in my early twenties or thirties.

Timberland says that AI tools help him focus on being a better artist, and that at the end of the day, he's the one pulling the levers, the one with the vision. But as AI changes the way music is made, labels, artists, and listeners are left wondering who stands to benefit from these tools and who has the most to lose. Today, on the show how AI broke through in the music industry, the sticky questions it raises, and where it could be headed Next, this is the big take from Bloomberg News.

I think AI was bubbling up for a while. Real prognosticators in the space sat coming and they were like, Yeah, this is going to be something where you can generate a song at the top of a button, and this is going to eliminate the need for commercial libraries. For example, you know the background music you hear on Netflix, or you license maybe catalog of songs and you can just

When you say someone's likeness their musical likeness, or their physical likeness, voice.

Musical stylings, sound essentially. So that's kind of more of the side where you see the labels and other rights holders kind of racking their heads about this, like what are we going to do.

One of the more high profile cases of this came in twenty twenty three, when an anonymous creator released a hip hop track called Heart on My Sleeve. It was generated using AI, but you can hear how it's designed to sound a lot like the artist Drake.

I can my eggs, and the song sounded like Drake and Weekend had collaborated. And when it came out, I think there was all this speculation like, is this a real song from Drake in the Weekend, Like we didn't know it was coming. It's a surprise drop. Certainly sounds like them. Eventually, when it was realized that it's an AI song, that just kind of started this whole conversation around AI in the music business because Drake and The Weekend are both signed to Republic, which is owned by

The song went viral, it even looked like it was on track to chart, but then it started disappearing from streaming platforms one by one. UMG wouldn't comment on it directly, but it issued a statement saying that training generative AI on music from their artists violated copyright law and that music platforms have a quote fundamental legal and ethical response ability to prevent the use of their services in ways that harm artists unquote, But some artists are embracing AI

Twenty twenty four was a really big year for AI.

Ashley spoke to Michael Nash, who's the executive vice president and chief Digital officer at UMG, about the second main way AI is being used in music production, behind the scenes to up recordings and in the studio by engineers who mix and master work. He told her about how UMG revived old music by the Beatles, a band the label represents and whose members aren't all alive, to re record the.

Story there is recordings in nineteen seventy seven by John Lennon were used to produce some singles in the nineties, but there was one track now and then which wasn't possible to use because the piano part was like kind of banging over the vocal and becoming intermingled in a way that the vocal track was not previously usable.

They used AI to extract John Lennon's vocals from this old recording that they had and essentially make it sound really nice and use it in the song.

So no part of that vocal was AI generated, but it was AI excavated.

It's now now needed for a Grammy.

Michael told Ashley that UMG is also interested in exploring another type of a music product. The third category of AI material that's cropping up in the industry content tailored to a specific listener.

There's an interesting component around hyper personalization, where the consumer can use the AI technology to customize the music listening experience.

Sony Music Entertainment, one of UMG's competitors, tried this last fall, the company worked with David Gilmour from Pink Floyd to offer fans a personalized AI track and artwork remixed from portions of the album.

So the idea really was, you're a fan, you want a special experience. You want to feel like you really own that album in a different way than anyone else, So you would get customized artwork based on your prompts as well as a song.

But whether there's consumer demand for this is another question entirely. Right now, the industry is in the spaghetti at the wall phase of AI, trying things out, seeing what sticks, and making case by case to decisions about what feels ethical coming up after the break, How companies like UNNG and musicians like Timbaland draw those lines and the legal

He's like going to the record store picking out samples, and when you get that right sample or that right record, it allows you to get more creative.

But Suno also lets users create whole songs. You just put in keywords, click a button, and generate a track. To understand how tools like Suno really work. Ashley and I gave it a test drive ourselves. Have you tried this before?

No, let's do it? Okay?

What should we make a song about?

Should we try to do? Like a Christmas song?

Definitely a Christmas song about New York and the beach, because I wish I was at the beach in the.

Style of Mariah Carey.

Mariah Carey, create a song. I'm so excited. In just a few seconds we had our custom song New York.

Noelle, Okay, let's play. It doesn't sound like Mariah Carey at all.

Christmas Tree, Sandy Grain. Christmas Tree is so bold and pretty, but I'm dreaming of Sandy Grains. It should be playing in the elevator.

That's exactly the thing. Like could this past for elevator music or podcast music?

Absolutely?

Yeah?

Like background music. I don't know. I want to believe that between the two of us we could write better lyrics for a Christmas song. I want to believe that, but I haven't actually tried. Definitely, wouldn't say Sandy Green.

Yes, I have never thought that sentence in my entire life. The way Timbaland describes how he uses technology like Suno is pretty different. He told me it would be hard to just auto generate a hit. That making music is still about a person's own creativity, something AI can only amplify.

I don't like to say the word train, but it is like a training. I'm transferring energies to a machine.

So do you have any concern that other artists might use Suno or a tool like it and recreate the Timbaland sound?

Just never going to give you me unless you know how I think you know everybody tastes ain't great. You know what I'm saying, That people are making stuff that they like to make music, but it ain't good music.

But UMG and two other major record labels have a deeper concern with the way Suno works. They've sued the company and a similar one called Udio, accusing them of

Honestly, we're just waiting for the courts to decide.

While they wait, UMG has started to make its own judgments about what it deems ethical use of generative AI. One example came this holiday season when the label released a new version of a Christmas song originally recorded in nineteen fifty eight.

So Brenda Lee's famous song Rocking around the Christmas Tree. We all certainly know it. They used AI to essentially train on her voice as a young teenager and record the song that sounds like her in Spanish.

Brand Lee's vocals were used as the training set to develop a vocal model, and then a Spanish language performance by an artist that was in the right vocal range for a match that provide a guide track, and then in the production studio, by merging the guide track with the AI voice clone, Wila, you have Brenda Lee performing Rocking around the Christmas Tree in Spanish for the first time.

She's an eighty year old woman. This is not her singing today. She did give her permission. They did this with her approval, but it is certainly her younger voice singing in Spanish.

To UMG, what's the difference between Heart on My Sleeve and AI generated AI assisted songs that they put out this year? What's the key distinction there in their mind?

UMG says they did not provide the permissions to release that song to train on Drake's voice or the Weekend's voice, that this had to have been trained on their copyrighted material, so therefore they do not approve.

But for some artists and companies, the fears around AI go beyond copyright concerns and their fears that pre date AI too.

There's kind of these bigger issues around how do you make money as a musician, as an artist, as a producer, how do I break through as an artist, how do I actually make money from streaming? How do I affordably tour?

AI makes some of those questions more pressing. If streaming platforms are flooded with AI generated tracks, for example, it could get harder for artists to break through the noise and get paid.

There's tons of tracks being uploaded a day. I think it's over one hundred thousand at this point now a day. The way streaming is paid out is out of a pie basically, So if Taylor Swift accounts for fifty percent of listening on Spotify, she gets fifty percent of the royalties. So it's taking away that market share and that money from legitimate arts.

Michael Nash from UMG says he doesn't believe a machine will ever be able to mimic the artistry or the emotion that goes into making a truly great song.

What fans really care about is the expression of human artists telling their stories about you know, their lived experiences, about people falling in love, about people having great friends, about people having a wonderful time on a Saturday night. That's what fans are really interested in. They're not interested in knockoff derivative soundlike content.

Whether or not listeners actually connect with music that's been touched by AI or even no AI was involved. The genie is now out of the bottle, and even Timbaland acknowledged that comes with trade offs. How might this have changed your process working on some of your earlier songs, Like would the way I are have sounded different if I had.

Existed, You know that's funny. I think I think it would have the same. It would just be the same way I'm using it now. It would just be more helpful and we would have put out way more music. The process has been way more faster. The reason why I'm glad that it wasn't back in because it wouldn't allow me to collaborate with people. If it was available backed in, it would have been like me in the machine.

This is the Big Take from Bloomberg News. I'm Sarah Holder. This episode was produced by Julia Press. It was edited by Aaron Edwards, Felix Gillette, and Dana Walman. It was mixed and sound designed by Alex Uguerira. It was fact checked by Audreyanna Tapia. Our senior producer is Naomi Shaven. Our senior editor as Elizabeth Ponso. Our executive producer is Nicole Beamsterboord Sage Bauman is Bloomberg's head of Podcasts. If you liked this episode, make sure to subscribe and review

Hmm.