As ChatGPT sets the internet aflame, will AI write the lyrics of tomorrow’s hits?

Welcome to the latest episode of the Music Business Worldwide Podcast, in which MBW founder Tim Ingham speaks to Dr. Maya Ackerman – the CEO and co-founder of AI lyric-generating platform, LyricStudio. The MBW Podcast is supported by Voly Music.

Bloody Elon Musk. Not content with doing beastly things to Twitter (like, erm, marginally improving the user experience), he’s also triggered the actual beginning of the actual end of the world.

Well, not directly. But Musk was one of the co-founders and funders of San Francisco-headquartered OpenAI, which is the progenitor of ChatGPT – the AI-powered online tool that can not only research and regurgitate online facts, but also weave that regurgitation in a variety of generated tones and styles.

Some tech experts are so impressed with ChatGPT, they suggest, in just a few more iterations, it will become a serious challenger to Google, and turn the business of online search upside down.

(The next chapter in this sci-fi novel: ChatGPT gains sentience, turns our own long-trusted devices against us, and subjugates the human race. But that’s probably still a few years away, so chill out, and, as Warren Zevon famously recommended, Enjoy every sandwich.)

The music industry is actually a little ahead of the curve on this topic. Because language-based ‘generative AI’ platforms are already transforming this business in a meaningful way – in the world of lyric writing.

LyricStudio, owned by California-headquartered parent WaveAI, produces original lyrics for songwriters in a style that mimics their own. In this sense, it’s ‘assistive AI’ – a human companion, a muse.

“When it comes to curing writer’s block, there is nothing as powerful as LyricStudio,” its website boasts.

One songwriter, FLOSS, aptly sums it up: “It’s like having a songwriting partner who has great ideas that you can either use or not use, but you cannot hurt their feelings.”

LyricStudio’s popularity is already mind-boggling: to date, it has been responsible for ‘assisting’ the creation of over a million songs, from over a million songwriters, musicians and producers. At least 15% of the people that use it, say LyricStudio, are professional music-makers.

One of those artists, rapper Curtiss King, released a No.1 album (on the US iTunes chart) last summer – with lyrics written/’assisted’ by LyricStudio.

On this Music Business Worldwide podcast, the co-founder and CEO of WaveAI/LyricStudio, Dr Maya Ackerman, discusses the future for AI and music… especially when it comes to lyric writing.

Ackerman has some powerful credentials: she is a professor of AI at the Computer Science and Engineering Department at Santa Clara University, as well as a singer, songwriter and music producer. She earned her PhD in Computer Science from the University of Waterloo, held Postdoctoral Fellowships at Caltech and UC San Diego, and has published over 50 peer-reviewed research publications.

We ask her all about LyricStudio and the moral and artistic quandaries presented by the use of ‘generative AI’ in music-making. She points out that, so long as everyone in her field behaves ethically – which they surely all will, right? – there shouldn’t be a music-biz-ending conclusion to this tale.

Listen to MBW’s Podcast interview with Dr. Ackerman above, and/or read an abridged version of our discussion below…

What in your view differentiates what you’re doing with LyricStudio from any other generative AI technology in the music space?

LyricStudio has been built in-house specifically for [the creation] of lyrics. It does not do anything else. It also writes in your style – that’s a highly unique feature of our algorithms.

Where do you stand on the argument that a tool like this strips the music industry of some of its humanity, or that it removes some of the humanity from the songwriting process?

That’s a really important question. There is a lot of buzz around autonomous creative AI systems – systems that [make] things from start to finish.

LyricStudio helps you write songs from start to finish; it’s actually designed not to [create a full song’s lyrics] by itself.

It helps when you get stuck. It has helped people finish songs that have been sitting in their drawers for years. But there is actually absolutely no way for it to function independently – and that’s by design.

To perhaps put listeners at ease, there’s absolutely nothing out there [in generative AI today] that works without human intervention, that does a good job at writing lyrics, start to finish, by itself.

I get the argument of generative AI being like a musical instrument within itself; a tool that musicians can rely on to either augment or improve the melody or lyrics that they are putting together. But it’s fair to say that there is a pretty high level of paranoia at this point in the music industry around AI music-making platforms. You only need to look at TikTok buying Jukedeck a few years back, and what ByteDance‘s intentions might be in making its own music on its platform. On balance, do you think generative AI going to be a plus or a minus for those artists and music rightsholders who currently do well commercially out of the music industry?

It’s complicated. There are going to be both pluses and minuses. It’s important for developers to admit that it’s not a purely utopian situation, and to be responsible in how they bring technology to market. Like any really powerful thing, AI is not inherently good or bad. It’s how we use it that’s really going to dictate [AI’s] impact in the world.

“AI technology itself has no goals, no motivation, no concept of ethics or humanity for that matter.”

The technology itself has no goals, no motivation, no concept of ethics or humanity for that matter. So I really hope that most developers, as this field grows, see that they should be bringing technology into the world that will help human creativity, rather than hinder it.

To be honest, the more that I hear musicians’ concerns [about AI], the more I end up taking their side, as far as needing to be careful, thoughtful and responsible in the way that you convey these technologies.

One of MBW’s biggest stories last year was about Tencent Music in China, who said THEY’D CREATED OVER 1,000 TRACKS on their platform that contained AI-created vocals mimicking the human singing voice. Some of those tracks had many millions of streams, one had over 100 million, and some of them replicated the vocals of deceased Chinese pop stars. Do you see this vocal-replication model becoming a trend in the ‘Western’ world?

Let me contextualize this a little bit from a technological development perspective. One of the unfortunate things as far as [today’s] musical AI systems goes, is this tendency to call [AI-created pieces of music] autonomous, when they’re really co-creative [i.e. involve significant human input].

The technology is just not there right now [in Dr. Ackerman’s view] to fully autonomously create vocal music. So what a lot of companies end up doing is they have human artists collaborate with AI, and then they essentially pretend that it’s autonomous, or at least they bury the fact that there was any human participation.

“in vocal technology, there’s definitely been much faster development in this domain for several Asian languages than English, although English is starting to come along.”

That said, [in] vocal technology, there’s definitely been much faster development in this domain for several Asian languages than English, although English is starting to come along. But very often with this technology, you have to tinker with it so heavily. So the AI kind of helps you make it sound like [a human] artist, but it remains a notoriously difficult thing to do.

That being said, these opportunities for co-creativity [on vocals] are really big. What I hope to see, as generative AI becomes more common [in music], is more admission of the cooperative nature of this [by companies making music with AI].

If [AI platforms are] designed in a way that genuinely nurtures human creativity and human ability, even independently from the machine, then you can see how it’s going become harder and harder for fully automated [AI] music to genuinely get ahead [of human-driven music].

Although I’m very impressed by LyricStudio, I also think it fits into this paradigm whereby there are great demands on the creativity of modern artists, who are sometimes expected to create and upload new music every week or two. I worry about the fact that those artists don’t take the time to be inspired. Or as an executive called David Joseph – via LYOR COHEN – put it recently, the idea that an artist is having to be ‘always on’ as opposed to ‘occasionally brilliant’. I fear that something like lyric studio facilitates more of the ‘always on’ mindset.

I’d love to contextualize this response by sharing a little bit of the history of our products. I’m primarily a researcher in artificial intelligence, generative AI, and have been for eight years. But while getting my PhD, I also learned how to sing opera, and I play piano. So I’m a classically trained musician.

When I wanted to write songs, I [often] found myself stuck. It was an almost spiritually frustrating experience, to feel that there is music within you that you can’t get out. Originally, when we created the technology that now powers both LyricStudio and [Wave AI’s upcoming music-making platform] MelodyStudio, it was done as a research project, to secretly to help me in my own [songwriting] struggles.

“What I find really exciting about our technology is that it can help to profoundly inspire people; it can help them gain a skill that helps them express what’s within themselves.”

There are people who use our systems to speed up their process – and I’m happy to be helping artists in any way possible. But that’s not the most revolutionary use case. What I find really exciting about our technology is that it can help to profoundly inspire people; it can help them express what’s within themselves.

While the fact that it’s AI is kind of cool and exciting at the moment, the long-term legacy of our products is going to be that they make musicians more creative. And it’s a difficult thing to believe, for some artists, because of the reputation that AI has, but these tools are really built to tap into human ability.

The users of LyricStudio have told me so many times of how deeply inspiring it is. Curtiss King, who had a No.1 hit on iTunes via an album that he made with LyricStudio, said that it’s like lightning in a bottle.

It enables you to capture this kind of inspiring moment, but also stay in ‘flow’ – to stay solidly in this creative space. You can get out this inspiration onto paper or onto your monitor before it fizzles away.

The LyricStudio site says the music made with the platform is 100% copyright free. That allows the artist to take credit. I feel a little unnerved by that because if you look at hip-hop, for example, suggesting that another artist uses a ‘ghost writer’ – which Meek Mill accused Drake of a few years back on social media – is a real slight. How do you feel about artists taking credit for the lyrics that LyricStudio creates? It doesn’t seem fair to me.

We gave this one a lot of thought. There are many reasons why LyricStudio doesn’t take credit. It’s the right move ethically speaking, because people use the [AI-generated] lines to get started [on their own material]. By the time a person gets to a place where they’re going to publish a song, they inevitably put so much of themselves [into it] that to say that the machine wrote it for them is definitely not appropriate.

Why does the machine get nothing [in terms of credit]? That’s because the machine is a different kind of creative entity. It is so easy for a machine to create ideas – it can do this all day and all night, serving millions of people – that for it to take royalties, for it to take credit, at a certain point is just greed.

“It is so easy for a machine to create ideas – it can do this all day and all night, serving millions of people – that for it to take royalties, for it to take credit, at a certain point is just greed.”

With ghostwriters, they get paid; a human needs to get compensated for that. And I understand there is some dishonesty there: You’re singing a song that someone else made. I understand why people feel uncomfortable with that.

But in the case of a machine being that partner that helps you move along, when you actually are kind of directing the whole show and making all the final decisions, and in most cases, actually creating a lot of material yourself? Any sort of assistive AI should not be keeping any of the credits.

I’m in two minds over generative AI. On the one hand, I see the downsides and the reduction of humanity in music. On the other hand, whenever an appliance in the kitchen is created that makes our lives more convenient, with the same results as a more laborious option, there is generational disapproval over the fact something used to be hard, with a soulful struggle, and now it is not. That applies to music: I remember watching a UK TV show where Bjorn from ABBA ADMITTED THEY USED RHYMING DICTIONARIES when they were writing all of their timeless pop classics. Is a rhyming dictionary generative AI? Perhaps it is and it should get a royalty! Anyway, final question: with the rise of ChatGPT and other platforms, do you think that 2023 is going to be a transformative year for the impact of generative AI on society?

Yes. I genuinely believe that 2023 is going to be the year that generative AI gets a lot of mass adoption.

This is so surreal for me, because I’ve been studying it for eight years, and I never even considered that I would get to live through it becoming the hottest thing.

With ChatGPT, I am so in love is what OpenAI has built. I have built it into my own workflow. I have it constantly open. I think it’s brilliant.

“I genuinely believe that 2023 is going to be the year that generative AI gets a lot of mass adoption for a lot of use cases.”

I’m definitely expecting a lot of use cases of GPT to become really big. There is a possibility that online search will be disrupted, in the sense that for a lot of things that people would have gone to [Google etc.] before, they might now go to a chat with ChatGPT – perhaps a future iteration of it, or perhaps Google will release their own [rival to it].

I’m super excited to see what 2023 has in store.