Producing Hit Songs in the Age of AI – Terms of Service with Clare Duffy

This is Terms of Service, I’m Clare Duffy. Today, we’re going to do a little thought experiment. Think about a song you’ve had on repeat lately or one you can’t get out of your head. Do you know the story behind that song? Any behind the scenes details from the artists who produced it? And what if you found out that song or even just the melody was generated by someone feeding a prompt to AI? Would it change how you feel when you listen to it? The music industry is yet another place where AI is changing the game. AI tools are poised to transform the way music gets written, recorded, and produced. And artists have mixed feelings about it. To better understand what all of this means for the future of music, for the people making it, and those of us listening to it, I talked to Harvey Mason Jr. He’s the CEO of the Recording Academy and the Grammy Awards. Harvey has created music with Beyonce, Aretha Franklin, Whitney Houston, Justin Bieber, Ariana Grande, and Elton John, among many others. He has a deep understanding of what makes a song a hit, and he has plenty of thoughts on what musicians need to do to thrive in this age of AI.

Well, Harvey Mason Jr., thank you so much for being here and doing this with us.

My favorite thing to do, we’re going to talk about music, tech, it’s good to be with you, Clare.

So you are a producer, a songwriter, but you’re also the CEO of the Recording Academy and the GRAMMYs.

I know, it’s weird, right?

What does that entail? What does it mean?

I’d been writing songs my whole life. I’d be in the studios since I was a kid growing up around the community. And I got the opportunity to serve the community as first the board chair of the recording academy and then the CEO. So very different mindsets, different hats, but at the end of the day, I love music. I love the people who make music and I really believe in the power of music and what it means to our world and society at large and just how we interact with each other. You could have the worst argument in the world, but if you go to a concert and you stand next to somebody and the music’s playing, you start dancing, you start singing, it just changes everything about your makeup. So that’s the reason I’m in the role, and that’s how the commonality between producing music and making it and sharing with people and protecting and serving others that create it is kinda how they tie together.

Got it. So this role, in part what you’re doing is advocating for musicians and artists.

100% advocating, serving, lifting, showcasing, highlighting, all that. It’s every day I wake up hoping to do a better job of that than the last.

And when did you sort of realize that AI was really going to be impacting the music world? Do you remember the moment that this dawned on you?

Of course, it’s like when you saw the lunar landing, or when you saw an assassination of a president, almost as that serious for creators. I remember being in the studio and somebody saying, have you seen such and such application and how it works around music creation? I said, no, I haven’t seen it. Pulled it up on the laptop, typed in a few things, and it started spitting out music.

Before that, I had seen it in other iterations, not quite as advanced, you know, maybe things that could enhance a vocal or things that can create a drum loop or a chord progression. But when I started seeing it make whole tracks and whole songs with singers and vocals and arrangements, it freaked me out. I really had to stop my tracks and say, what does this mean?

So just to sort of set the stage here, will you give us the cliff notes on what it takes to make a song? Like what kinds of human inputs are required to get something that we hear on the radio?

It’s countless decisions and choices and a series of compromises and trying to get what’s in your head out can take a lot of different forms. Some people sit at a piano, start playing chords and then start singing melodies and start thinking of words. Other people use computers, different digital audio workstations to record MIDI information and build tracks piece by piece, but it’s thousands of decisions, everything from what notes you’re playing, what chords you’re paying, what’s the tempo? What’s the key? How many instruments do you have? What’s the drum pattern? What’s base pattern? How do you fit the lead vocal over the track? What are you saying? How are the melodies flowing? What key, you know, there’s so many critical decisions that go into making a song. And up until this point, you’ve had to had some real specific musical or traditionally musical concern talent to be able to do all those things. You had to be to play an instrument. You had be able have an ear for what you liked. You also had to, most of the time, at least for me, I’m hearing something in my mind that I’m just trying to figure out how to get it out. So that comes from somewhere else and then we don’t know where that comes from or at least I don’t. And to be able to figure out how express that is I guess up until this point what’s considered a traditional music creating talent.

And based on the tools, the AI tools that you’ve seen, how are those different? Like, what are those changing or replacing in the process? Potentially all of that?

You nailed it, it’s replacing all of it. It’s replacing the composing part, the coming up with the chords, it’s the replacing the production of arranging and deciding the layering and what kind of instrumentation and rhythms go well together, which ones don’t. It’s replaced melody creation, lyric creation. It has gotten to the point where anyone you know, or anyone that knows how to use a laptop in the most basic form, typing a sentence, even if you’re on your phone, texting. Those people can create music using AI.

‘Do you have a sense of what it takes to make good AI-generated music? Like, surely anybody can type into one of these programs, but is there a difference between somebody who knows how to use them to make something that sounds like it might pass for human-created music? Or is it really just that easy that anybody could do it?

In some instances, it’s that easy. If you get lucky based on what your prompt is and what it gives back, it has been known to create some really, really interesting, compelling pieces of music. The better you are at producing and writing songs and being a creator, to me, I believe you can have an effect on the output of the AI. You’re generating pieces, you’re giving it feedback, you’re connecting things to other things, and the way the software’s progressing, you’re gonna be able to create sections. And then it comes down to taste. Does this go well with this? I don’t like this. Let me try five or six others. Connect these. I don’t like this lyric. I don’t like this melody. And be able to interchange and exchange pieces of the song. That’s probably the next iteration of how AI is going to be going forward right now. You kind of give it your best shot at a prompt. It kicks something out to you. And a lot of times you can go back and change it. Let me speed it up a little bit. Let me slow it down a bit. Can you add a saxophone? These are the types of prompts you can do. So to answer your question, it’s a little relative because we need to have a conversation about what do you think is good music and how do you quantify that?

What kinds of conversations are happening in the industry right now about this? Is it mostly anxiety? Is there any excitement?

‘Yeah, there’s a lot of excitement and it depends on who you talk to and it also depends on the day of the week. I think people swing wildly from complete panic and terror and fear around AI replacing everyone and disrupting the entire industry to AI is going to be the biggest windfall and boom for the industry since maybe streaming because they see the potential of AI as far as the positives, making more music and allowing more people to create music. I think is a great thing. And that’s why I’m excited about music. I am not an anti-AI person, I understand that AI is here, I understand how powerful it is, and I see the value in allowing people to be more creative, to come up with things maybe they didn’t think of originally, to be able to try things and demo a bunch of different ideas before they land on something they really, really love. And being an arbiter of taste is going to be really, really important. So that’s some of the positive stuff. The scary stuff is, what is AI learning on? Like is it being trained on copyrighted material that, you know, maybe I wrote a song 10 years ago and now people are using that to make their own songs. That’s got to be clarified and cleaned up with some legislation or at least some guardrails.

Is it going to replace human creativity is a big fear. We hear a lot of that coming around from both musicians and songwriters and singers. And also, can it use your personal name, likeness, style, sound, voice? If I’m a producer and I produce a thousand songs, and I have a certain way of producing, can it listen to all my music and then say, hey, I want a song that sounds like Harvey Mason Jr. produced it, is that protectable? And so those are some of the conversations that I think everyone’s having from individual producers and writers all the way up to the corporate label owners.

Obviously, and you mentioned some of them, we’ve seen technology come in and sort of evolve, transform the music industry before. There’s autotune, there’s a synthesizer. Does this moment feel fundamentally different? Is AI different from those other technologies?

I think it is because I came up in a different time and I created music in a different time. I’m sure the generation that’s using it now thinks it’s the same as any other development or disruption in the industry. You know, piano players were scared to death of the piano role. You now, the player pianos where you put the thing on would play by itself. They said, oh, it’s gonna get rid of all piano players, you know, and that was a real thing. And then, when the synthesizer came out, everyone was afraid that there’d never be any more orchestras because it could simulate an orchestra or a violin player. I’m sure right now, to us, it looks like AI is so dissimilar to any of those other analogs, but I think in the future, as we start to wrestle with what it means for us as creators and we start adapt to it, because for me, that’s what I fall back on, creators and people who make art and make music. We’re very, very perceptive, and we’re also resilient. We take what we have and we figure out how to use it, and we make great new stuff with it. So I do think that once it settles in, that this is where music is going, we’re gonna figure out how to do it the right way. I believe in the work we’re doing at the Academy to make sure human creativity is protected, and human artwork that’s been copyrighted or protected is continuing to be protected. So there’s a few things that have to happen, but I’m still optimistic that this can be a way of expressing ourselves musically in a way that had not been done up to this point.

Do you think that there is a real risk that human artists get replaced by AI, or that it makes an already competitive industry that much more competitive? It sounds like maybe that’s part of what’s fueling the work that you’re doing right now.

It’s definitely going to make it more competitive. I don’t think there’s any question about that. And I’m already pretty positive it’s the most competitive industry in the world with so many millions of people making music. As far as if it’s going to replace human creativity, that’s really the struggle for me. Because I don’t believe human creativity can ever be replaced. I believe what we have lived and experienced in our lives, how we’ve grown up, how we been domesticated, how been in relationships, our interpersonal interactions make us who we are and we express that through our songs. Having said that, I’ve seen AI copy what humans can do. So, it’s not going to be exactly human, but I’ve seen it learn pretty darn well how to emulate or even predict what somebody’s going to think or say and then put that on a song, lyrically, emotionally. So, I don’t have the answer. We can be replaced. I like to believe we can. Because I think we’re inherently very special and we’re emotional and we have things that I don’t think any computer can ever do. But I do believe we have to be competitive. I think that we have figure out how to use this technology. I don’t think we should be hands off and say, oh, I don’t like AI, ban it. I think better get invested, better get our hands on it, figure out to use it and put it to use for our creative community and make sure that it’s protected. Copyrightable, it’s monetizable. Make sure anything learning backwards is taking care of the people it’s learning from and anything we make forward is getting approval or royalties of the right people. So I’m optimistic but scared at the same time.

Have you started making music with AI yourself?

I use AI all the time, I will say I use it in business, I use it in music, I used it in writing, and I’ve tried to adopt my own advice, which is get into the AI and see what it does, see what the capabilities are, how can I use it, because what’s going to differentiate one songwriter from another, one producer from another is how we interpret the technology and how we utilize it to make something great. And I said it earlier, but I do believe taste is going to come into play. Some expertise about music and knowledge about how things go together will help, but AI does a lot of that for you. So it’s going to come down to making sure you have a bit of a finger on the pulse of what consumers like, and if what you’re making happens to align with what people like, that will put you in a place where you can be successful. So the earlier you get your hands on it and figure out what it can do, the better.

‘So I know the Recording Academy has started to sort of evolve its thinking around the GRAMMYs in particular, music that includes AI components can win awards, but not for those AI components, if I’m understanding it correctly. Do you imagine that we might one day see an AI-generated music category at the GRAMMYs?

I won’t say never, but I don’t imagine we would, just like we don’t see a category at the GRAMMYs that says best synthesized record and best organic instrument record. I believe if we can figure out all the legal issues and some of the other challenges we have with AI, I believe that AI will be used as a tool to create more music that hopefully The Academy can continue to honor, honor accurately. We’ll still have different genres and different styles of music, but AI will be the tool that helps really creative and talented, passionate people make their next version of art.

So, you’ve touched on this, but I want to talk about the protections that artists are going to need going forward. We’ve already heard these concerns from artists that their music is being used to train these systems that could maybe replace them at some point. Talk to me about the NO FAKES Act. I know that you’ve been on Capitol Hill advocating for that, why do you think that that’s the right approach?

I’ve been on Capitol Hill more than advocating. I’ve jumping up and down, waving my arms, saying, we’ve got to pay attention to this. It took me a long time to perfect my craft and get to the point where anyone wanted to hear my music, let alone liked it. So I’m not comfortable with just giving that to somebody and letting them do whatever they want with it and learning from it and creating new material off of that. So, personally, I am invested in making sure that this goes the right way. And the Academy, we represent 23,000 plus members that are different phases of the industry, whether it’s engineers or writers or artists. And we all believe the same thing, we have to protect our intellectual property properly. And so that does involve creating legislation or advocating for legislation. We’ve got some success on the state level in California, Illinois, and Tennessee, where we passed something that was called the Elvis Act, but it really is, how do we protect the name, likeness, and voice of singers and artists? And so that’s a bit of legislation we’re working on. Federally, it’s called the NO FAKES Act, as you said. We have some support. I don’t see a lot of dissension for it. It’s pretty much been supported across both sides of the aisle. And I hope that it can get introduced this year because I think it’s an important landmark for us to be able to say, AI, you cannot take a person’s voice and use it any way you want. And somebody else can’t prompt a voice to sound like an artist. That’s not okay.

Do you imagine that, assuming that legislation passes, are there artists who are interested in saying, I’m willing to let other people use my voice as long as I get some credit or some money from that?

It’s great. Every artist should have the opportunity to do with their voice what they want, not what somebody else wants. If you want to have AI use it, you should be able to, as long as you have the things that you need, approval, remuneration, credits, those are the kind of things that we have to make sure. Because right now, I can make a song, I could make a vocal version of you, Clare, and do an interview in your voice and put it on a podcast. And you can say nothing about it.

I can make a song that sounds like my favorite artist and write it myself, sing it myself in my bad singing voice and put that artist voice over mine and put it out.

I don’t have to ask them, I don’t have to credit them, I don’t have to pay them. So this is why the NO FAKES Act is so important.

We will continue this conversation after a short break.

Okay, something I’m personally really curious about, we know that there is so much power in learning to play an instrument, even if you’re not a professional artist, it’s this emotional sort of expression, it can create community. Do you worry that with AI people might stop wanting to learn how to play music even just for themselves? Do we lose something if that happens?

Personally, I think yes, we do. Playing an instrument is important. The development of what that means to you and the creative side of your emotional personality, the development in your brain of what music means to someone who learns how to play an instrument is proven. Having said that, the people who used to criticize DJs and people who played turntables and then people who now tour the world playing music out of the laptops They would say oh those aren’t musicians those aren’t creators and clearly they are and clearly they’re incredibly amazingly talented at what they do. So my hope is that we can continue to have people who play traditional instruments, guitar, the piano, the bass, the drums many many others that I’m not mentioning, but also have people who create music in a new way and I’m sure young people who come up on the laptop prompting AI, aggregating different ideas, putting things together, compiling sounds, and making music, they’re going to believe that they are musicians. And to me, it’s a fundamental question that maybe we can ask your audience to consider, is anyone who makes music a musician?

‘Right. I’m fascinated by this. What do you think? Like, if you are just typing a prompt into AI and music spits out, maybe we won’t ask if that person is a musician, but is that the same as human-created music or is it something else?

‘No. That’s not the same as human-created music. It is music.

But who’s to say what’s the right music, or the best music, it really depends on who’s listening. And who do people want to listen to? If something sounds great and it resonates with you as a listener, as a consumer, do you really mind where it comes from? That’s a question, a philosophical question that I don’t have the answer to. But I believe people less and less care about the technical side of how the music was made. And we saw the example two years ago when the first kind of bits of AI, there was a song that came out that kind of took the voice of two artists and put it together. It got a couple hundred million streams in two weeks.

This was Drake and The Weekend.

Right and some of it was probably curiosity, but also some of it was people just thought it was a cool song. So it’s gonna be something we’re gonna have to contend with. I definitely don’t have the answers. It’s a very complex subject. It’s gonna play out pretty quickly because the technology’s here and it’s evolving so fast.

What advice do you have for artists who are interested in experimenting with this technology but also don’t want to lose too much of their voice or their control over the process?

I would say make your art however feels right to you. I don’t think anybody has to do one or the other. Personally, I do believe that AI is gonna be involved in the future of creativity, whether it’s music or film or photography or image generation, I believe that. But if you choose to not do that and use it, then I support that because making music is the purest form of expression. And if you feel like you need to express yourself on a banjo you should do that. If you want to do it on a computer, you should do that. To me, I don’t discriminate. If you can make great music and if you can say things that are important and say things that move somebody either to, you know, love someone else or to cry or to have a party or to wake up and dance, I’m not so precious about how you did it. I would like to think that traditional musicianship can continue because I think it’s important. My father was a musician. I’m a musician. So, I like to think that that’s a thing, but also I’m not gonna have my head stuck in the sand to think, that’s the only way we have to make music.

Do you expect that we’ll see a surge? I mean, obviously it’s already such a big business, but in more people wanting to go see live music in person?

I hope so. I think it could. The more that you don’t know where the music comes from or who created or how it was created, the more maybe people will wanna go and see the actual person performing and the manifestation of the creativity you’re hearing on a streaming platform. Now you can experience and witness it firsthand. I hope there’s a huge appetite for that. I know I love that. There’s nothing better than seeing an artist just pour their heart out on a performance. I mean, what beats that?

As a sort of groupie for my husband and I followed the Tedeschi Trucks Band around.

Oh, wow, nice.

And it’s like, yeah, you just can’t beat seeing 12 artists on the stage performing live, in my mind.

I agree.

But are there other things that individual music consumers, music lovers can think about if they want to continue supporting human artists? Maybe regardless of how they’re producing the music, but if it’s important to them to continue to support the humans that are making it.

‘The time that that question’s gonna be even more appropriate is when we start seeing AI artists. Right now we’re seeing AI-generated music that humans generally are singing. When we start having AI artists, then we’re gonna have to make a decision as consumers and listeners, do we wanna support computer art or do we want to support human art? And at that point, it’s just making that decision. There are producers that are right now talking about signing, creating, I guess I’ll say creating and signing, AI artists. Absolutely. You can go across your social media feeds. There’ll be producers, big producers that I love and respect that are creating AI artists and I can’t blame them because the ability to write a song now and have it performed by somebody that you have created and detailed and you’ve manipulated to sound just the way you want it to sound is really cool. It’s what we all try to do as producers, whether it’s an artist that assigned us or an artist that we’re working for, we’re always trying to get the music to sound just the way we want it. So to be able to do that with an AI artist is powerful. So that’s gonna happen. And at that point, it’s gonna come back to that question, do you just listen to stuff you like and don’t concern yourself with where it came from or who did it? Or do you try and support that human creativity? And I don’t know what’s gonna happened. We’ll see.

Is that like in the next year we could be hearing music from AI artists?

You’re hearing music from AI artists probably already.

‘You might not know about, but you’ll start seeing more high-profile artists with names and pictures and album cover artwork and things like that that producers will get behind and labels will probably get behind. So I think it’s going to happen.

‘What’s one thing that you hope people keep in mind the next time they hear or see something about AI-generated music going forward?

I would hope when people hear something about AI, it makes them realize that there’s still got to be somebody making creative decisions. Right now, I think the word around AI is that it’s just replacing all human creativity. My hope is that AI and the technology around AI, generative AI for music, is that there still going to be the need for humans who have a musical ear or talent for gaging what people will like, or a creative sense about what’s great and what sounds exciting. My hope is that those people can still thrive. So as you’re listening to this conversation on AI, let’s try and remember in my ideal world, there’ll still be somebody that’s a producer or a songwriter, a tastemaker, somebody who’s using that technology in a way that’s different from everybody else, creating something that no one’s ever created before and expressing themselves in a new way that’s maybe not any better or any worse than the way we used to express ourselves. We’ll see.

Do you have a message for the AI companies about how you’d like to see them act going forward when it comes to this issue?

Yeah, I would say it sure would be nice if we could remember humanity, remember the people you grew up listening to their music when you were nine, 10, 15 in high school, going through your life experiences and living through the soundtrack of those humans that made that art and remain respectful of those people and make space for them to continue to be able to own and control their copyrights. And also make sure that we’re doing it the holistic view of what making music is, like let’s be sensible, let’s think about not just how fast we can crank out music but also how special we can make things and we can allow humans to express themselves. And I tend to think that the people running these platforms, I met with one of them last week, he’s a music person, he loves music, and he used to sing and…

Can you tell us who it is?

No, I’m not going to say who it is.

But he loves music, so I believe they’re trying to find a way to make more music and allow more people to make music. I just hope we can remember at the base of all of it is human art, human creativity.

Well, Harvey Mason Jr., thank you so much for doing this. This is a really fascinating conversation.

Yeah, we could go all day about this, it’s not stopping.

Yeah!

‘So whether you’re a musician, a producer, or just an avid playlist maker, here are some things to keep in mind. Whether you were aware of it or not, you’ve probably heard music by now that was generated, at least in part, by AI. Studios are already starting to create AI artists, so you’ll likely hear more and more of this. Still, Harvey says that AI can’t replace the spark of real-life creative collaboration. Or the human ear for what actually makes a piece of music compelling. At the end of the day, human decisions are still at the heart of a good song. As this technology continues to evolve, musicians do face a growing risk of their intellectual property and even their actual voices getting used without their permission. The proposed NO FAKES Act is one piece of legislation that’s trying to help artists protect themselves by holding companies and individuals liable if they create an unauthorized replica of an artist. Thanks for listening to Terms of Service. I’m Clare Duffy. Catch you next week.