AI-generated music comes to adland, here is the impact it will have

Then, in a flash, you remember that you have the power of artificial intelligence (AI) at your fingertips.

You pull up a generative AI model on a web browser, or maybe on a mobile app. You enter a handful of text-based parameters into a simple search bar: “ethereal,” “new wave,” “jazzy,” and “old school,” and – after a minute or two – voilà! The machine spits out the perfect song, complete with slightly modified alternate versions for you to choose from. And it’s done so at a laughably cheap price. It might have even been completely free.

It may sound a bit far-fetched, but anyone who’s played around with a generative AI model like ChatGPT, Midjourney or DALL·E 2 knows that this technology is already capable of some stunning feats of creativity – or at least, what looks to us like creativity. With just a few prompts, and perhaps a few rounds of tweaking, these tools can create content that, at first glance, could pass for the work of a moderately (and, in some cases, a very) talented human being. Many brands now have their sights firmly fixed on this emerging technology and its potential to completely revolutionize marketing.

And very soon, AI-generated music could emerge as the next big thing – pushing the generative AI revolution to new heights. In fact, it’s already arrived. You just probably haven’t heard any excited murmurings about it yet because it’s still in its infancy, a technological baby that needs to mature before it’s unleashed into the world.

Still, there are several companies (of varying sizes) that have already established themselves in the emerging AI-generated music space. For example, OpenAI – the company behind ChatGPT – has a platform called MuseNet. IBM has Watson Beat. Google also has a generative AI platform for music that has yet to be released to the world (more on that below). And there are many additional players who don’t (yet) have household names, including Amper Music, Aiva, Boomy and Soundraw.

Agencies aren't clamoring for AI-generated music, yet

At present, AI-generated music is more trouble than it’s worth, says Wunderman Thompson executive director of music and audio Paul Greco, “There are so many nuances that we could be doing this all day if we’re just loading keywords into the machine,” Greco told me over a Zoom call, his own complete music production studio on full display in the background. “Whereas if I’m working with a [human] composer, it’s happening fast … It’s one-on-one, we can try things immediately.”

Greco also shrugs at the quality of the “one or two” AI-generated music platforms he’s experimented with. “It sounds perfectly fine in a lot of ways … [but] I haven't heard anything that has blown me away yet.”

“Yet” being the operative word. AI-generated music may not currently hold a candle to Beethoven, The Beatles, Biggie Smalls or Beyoncé, but for how long? Let’s not forget that large language models, though they’ve been around for decades, were not particularly impressive until the past couple of years. In November, ChatGPT was released by OpenAI and the world collectively gasped at its capabilities and wondered: Is this machine actually intelligent? How much longer before the next ChatGPT rolls along and large numbers of people wonder: is this machine actually sentient?

The point is: AI can learn and improve; that’s kind of the whole point. And though these are still early days, AI could very well improve not linearly but exponentially, jumping from zero to one hundred virtually in the blink of an eye – meaning that a model that’s not-so-impressive one day could be an absolute world-changer the next.

And it’s worth remembering that AI-generated music doesn’t need to be as good as the GOATs to be good enough for widespread adoption across the marketing industry - it just needs to be better than the average human musician at creating catchy tunes for ads. If this were to happen, Greco believes that it would probably push out the less-innovative human musicians. “It might thin the herd out … [and] push those lower-tier folks out of the game entirely,” he says.

It could make these upper-tier musicians think: “‘I've got to do better. I can’t be complacent. I go to make sure I’m better than Google so people will pay for me,” says Greco.

There’s a reason Google is cited in this hypothetical scenario: Last week, it was revealed that the tech giant has developed a sophisticated generative AI model for music called MusicLM (the “LM” presumably standing for “language model”). But the company is reportedly waiting to release it until it can iron out some serious issues regarding copyright law – namely, ensuring that its AI model doesn’t steal the music of human artists. There’s probably an altruistic angle to this decision (namely, wanting to protect the rights of artists), but it’s also a simple matter of self-preservation: The launch of a generative AI model which steals music from human artists could land Google into some serious legal hot water.

But what about those pesky copyright laws?

Copyright protection is, of course, an increasingly salient problem for generative AI models across the board. Earlier this month, three artists filed a lawsuit against Midjourney and Stability AI (owner of the generative AI model Stable Diffusion), claiming that their AI-generated art platforms are plagiarizing the work of human artists. (Regarding the lawsuit, a spokesperson for Stability AI told The Drum in an email: “The allegations in this suit represent a misunderstanding of how generative AI technology works and the law surrounding copyright. We intend to defend ourselves and the vast potential generative AI has to expand the creative power of humanity.”)

Fortunately, there are some working in the AI-generated music space who are giving the issue of copyright protection serious thought. Alex Mitchell, CEO of Boomy, says that his background as a musician has shaped the philosophy of his company: “I was a scared artist who saw [generative AI] eight years ago and was like, ‘we need to make sure this happens in the right way,’” he says. “And to me, the right way means respecting copyright.” Boomy, according to Mitchell, uses no copyrighted material.

Marketers would be wise to take note. A generative AI model that can produce completely original music could, as we started to explore earlier, come in very handy for many marketers who spend much of their time searching through stock sounds. Mitchell says that Boomy’s user base probably includes “hundreds of thousands of marketers.”

Then there’s the cost factor. As Greco alluded to above in his example of the fictional musician hustling to outpace the AI, many brands are likely to turn to AI-generated music in the coming years simply because it’s cheaper than paying a human music producer. “I would encourage marketers to see the opportunity of [AI-generated music] not as one where costs are driven close to zero, but rather one where costs that would traditionally have gone somewhere else can now be redirected in other aspects of the campaign, [enabling them to] tell the best version of the story they want to tell,” Pierre Barreau, CEO of Aiva, told The Drum in an email.

“A practical example: If using generative AI has reduced your music creation costs by a factor of 10, you can allocate the funds you saved towards hiring a performing artist to perform some aspects of the song, or jam along with it.”

Bottom line: AI-generated music is a fairly young technology, but when it fully breaks into the zeitgeist – as ChatGPT did in late 2022 – marketers of all stripes should be prepared for the sudden and sweeping changes that could transform their industry.

For more on the latest happening in tech, sign up for The Drum’s Inside the Metaverse weekly newsletter here.