ai music image

Music AI tools are, generally, powerful and enabling: unlocking artists’ creativity and opening up opportunities for new, innovative forms of music. But of course, it can also be used for duplicitous reasons – and between those two extremes are a host of fuzzy questions that now need to be wrestled with.

For instance: you may have seen on social media a number of examples of AI-generated sound-alikes of famous singers performing songs by other artists. One of them, a cover version of “Drake” performing Ice Spice’s song ‘Munch’, caught the attention of the Canadian rapper himself, who took to Instagram to call it, tongue-in-cheek, we think, “the last straw”. At the moment, these AI covers seem to be slightly mischievous experiments, but lurking beneath it all are some complex issues. From an artist standpoint, there’s the simple issue of ethics. Having your voice cloned might feel like an intrusion: suddenly they have no control over their voices, and words are being put in their mouths.

That lack of agency – over something as fundamental as your voice – might be disconcerting enough, but the potential for intentionally harmful fakes is immediately obvious and plausible. It could be a powerful form of misinformation: hearing “Drake” sing Ice Spice’s saucy lyrics might raise a smile, but an AI-generated “Drake” singing racist, abusive, or even overtly political lyrics is a very different issue. As the old saying goes, a lie is already halfway around the world while the truth is still putting its shoes on.

It’s not just major label legal departments that are bracing themselves for some tricky decision-making: even assuming that an AI cover version is clearly labeled as such and its creator doesn’t attempt to make money from it, what happens if the public decides that they overwhelmingly prefer Rhianna singing “Cuff It” instead of the original Beyoncé-voiced version?

David Guetta recently slotted the “voice” of Eminem into a track he made, although he didn’t release it. This hints at a good, interesting future use of this kind of technology: while an Eminem x Guetta collaboration is eminently possible in real life, what if he could use the voice of a deceased artist, with the approval of their estate, and with an appropriate credit and royalties system in place for the use of it? There are many possible nuanced outcomes here: this approach could result in some exhilarating and previously-impossible new music (Freddie Mercury’s voice on a David Guetta track sounds all too possible, for instance). It could also mean a future where voices that are currently considered de rigueur become a kind of standard tool, like how the sound of the TR-808 drum machine has formed the backbone of dance music for decades. Do we as a music listening community want new voices, or more of the ones we already love?

The recently-launched Campaign for Human Artistry, which is trying to set guidelines for the use of creative AIs, and counts a number of large international music industry organisations as members, listed seven principles that it wants AI music platforms to abide by. One of them was around transparency: “Stakeholders should work collaboratively to develop standards for technologies that identify the input used to create AI-generated output.” This principle is possibly aimed at ensuring that songs generated by AI disclose which songs the AI in question was initially trained on, but also applies in these voice-replication instances too. Should these songs not be clearly flagged so that we, the listeners scrolling through a fast-moving landscape of endless content, are able to instantly know when a song is something that a human artist has created and approved – and when it just sounds or looks like them, but was actually AI-generated?

The big rightsholders are viewing AI-generated music cautiously for a number of reasons, and are particularly concerned about the provenance of that music, with UMG apparently flexing its muscles to stop people from training AI models on music without a license. Throwing unofficial AI-voiced covers into the mix doesn’t make things any less complicated; but it does show that AI is already reinventing not only how we make music – but how we think about it and where the artist’s value lies in it, too.

Music Ally’s next Learn Live webinar will help you understand what’s required for artists to thrive in new international markets!

Joe Sparrow

Joe SparrowEditor

Editor, Music Ally

Leave a comment

Your email address will not be published. Required fields are marked *