Imogen Heap Talks AI Music: ‘Anyone Who Has Spent 10,000 Hours Perfecting Their Craft Will Always Have an Edge’

Imogen Heap has always been known as an innovator in the music industry. The British singer, songwriter, producer and technologist has been experimenting with cutting edge tools to push her creativity forward since she first began releasing music over 25 years ago.

Now, as AI music continues to make headlines in the music industry and infiltrates the songwriting process, Heap is working on ethical ways to incorporate it into her own work. Recently, she released the song “I AM___,” a 13-minute epic that featured a collaboration between Heap and AI.Mogen, her self-trained AI voice model, and by collaborating with her digital self, Heap forces listeners to consider big questions, like the nature of art and self-identity.

She’s also working on a company called Auracles, which recently announced a partnership with SoundCloud designed to create a verified digital ID for musicians that, in the age of AI, helps them track their music’s uses across the internet, grant permissions for approved uses of their work and create “the missing foundational data layer for music.”

To talk through how she’s using AI in a responsible and creative way as well as the 20th anniversary of her seminal album Speak For Yourself, Heap joined Billboard‘s new music industry podcast, On the Record w/ Kristin Robinson, this week.

Below is an excerpt of that conversation.  

Watch or listen to the full episode of On the Record on YouTube, Spotify or Apple Podcasts here, or watch it below.

You’ve been watching this space for a long time, but when do you feel like you noticed a shift when everyone else started paying attention?

I feel like the silver lining, really, we definitely are at that place where it’s very confusing right now, and we do need some clarity, but we are, we can do it. There are tools, and everybody wants it. So I feel like the silver lining, really, of this dark cloud that’s on my seat as AI music taking over, is that we are going to get the data layer of works, we’re going to get this complete data layer of works because people will want to prove not only that they’re human, but they want to go there. They want to actually say, “No, I’m human and these are all my works that I actually contributed to.”

Hallwood Media is the first music company to be open about signing so-called AI artists like Xania Monet and imoliver. Do you think that major labels will start signing AI talent, or talent that uses AI very heavily, in the next few years?

I think a lot of things when you say that. I feel a lot of major labels are signing music that sounds AI generated to me anyway, it’s just like, ‘oh, that just sounds like that last thing and that last thing’ — nothing’s changed. So, it wouldn’t surprise me. I thought that that would happen, and wherever there’s money, obviously they’ll go if they think they can make money out of this artist.

You recently released a 13-minute song called “I AM__” which featured you singing alongside AI.Mogen, which is essentially your own AI voice model. Why did you decide to collaborate with yourself, and what were you trying to say artistically?

It’s kind of a very long winded, kind of silly way to do things, but I did it as a statement. I did it to rile people up, I suppose, and just be able to have this conversation. Because the song, in the beginning, it takes you through this journey of what I’ve experienced over the last four years where I started to think about, ‘Who am I? What am I? And this ego, what is it?

And then I started to think about AI. Because what is AI? What can I feel that AI can’t feel? The noise section in that song is like the annihilation of my ego —I’m not saying I have no ego now — But then, after I needed a section for after care. I wanted to explore this idea that AI is our child. AI is something that we are raising together, as the as the mother and father. Right at the end, there’s the voice of AI.mogen. I wanted it to be an AI voice, even though I had to sing everything. The way I made this is I sang all the parts, and then I put it through my AI model so that the model of my voice is then singing the words. It’s like changing the sound on a on keyboard.

Yeah, so like a voice filter over your own performance, but this filter is also you in a way?

Yeah, and I wanted to trick people. After this quite traumatic noise section, they would feel something, and that voice would not be my voice would be the AI agent’s voice. I wanted to create a discussion. I wanted to, you know, show people that we already don’t know the difference between AI and human, but does it matter? I do say in the title, it’s AI.mogen, but it was all ethically sourced. It was all done in the best way possible. And it’s my own voice, and I didn’t use any, I didn’t generate any music.

Of course, some people already canceled me for, you know, even saying that AI is in my music. So many people have said: ‘You use AI to generate the song?’ I was like, ‘No, I did not. I wish I could, because it took me four years to do it with 100s of hours,’ but the point of this is everybody is fearful of it, but we can still feel. And what is art? Art something for someone, not to somebody else, but if a sense of it, it speaks to you and it makes you feel something does it matter if it makes you recognize something in yourself? I mean, essentially, or AI is generated from human.

I love that you’ve been able to make this model that is totally within your control. I think the thing that gets scary is when anyone can create another person’s voice model online. I think such an important part of being an artist is having the taste and the curation to know what you want to say and don’t say. In an age of AI, it feels like you’re losing a lot of control over yourself. Do you have fears around that?

I mean, we already have lost control. People basically, you know, say that we’ve written something when we haven’t, and they don’t credit us when we have been a part of it. But that’s still very much less. I do think it’s gonna, really, it’s just gonna, it’s gonna force us into creating something that will make sense of what we have already and for the future, so that we can put a flag in the sand as humans and go, okay, up to this point, it was human generated. Again, I think it just comes down to this core missing piece that we don’t have which is an ID layer, like, identifying home for each individual.

AI almost completely lowers the barrier to entry for making music. As a trained musician, I’m wondering what your thoughts are on that?

Why not? Anyone who has spent 10,000 hours perfecting your craft will always have an edge. If you generate anything off these services right now, you’re just going to sound like 99.99% of other people who did that too, but if you have an edge, if you have a real something there that connects with people, you use these models differently. But, yeah, this helps everyone move forward. I don’t have problem with that at all.