Another Fake Drake? AI is decoupling artists and their voices (and there’s no coming back)
Photo: Karsten Winegeart
Drake has had perhaps one of the most publicly up-and-down relationships with generative AI. There was the AI Drake music generator, which he was dead set against. Then there was ‘Heart on My Sleeve’, the AI-generated mix using the voices of Drake and The Weeknd by a creator known as ghostwriter, where Drake again expressed displeasure at the use of his voice. Yet his tone seemed to have changed during his recent public beef with Kendrick Lamar, during which Drake used AI to include the voices of Tupac and Snoop Dog in ‘Taylor Made Freestyle’ (which was ironically taken down after Drake failed to secure permission from the late Tupac’s estate). He also recently sampled 'BBL Drizzy', a parody song with lyrics written by comedian Willonius Hatcher, musical elements generated by AI, and remixed by Metro Boomin, for his own song 'U My Everything'. He had to legally clear the use with Hatcher – possibly the first example of licensing for an AI-generated song.
Now we have a new chapter, with the mysterious drop of ‘Wah Gwan Delilah’. Released on Soundcloud by fellow Canadian artist Snowd4y, the cover of Plain White T’s 2006 indie hit ‘Hey There Delilah’ features Drake’s voice quipping about ladies and Toronto landmarks. It is unclear whether Snowd4y actually collaborated with Drake, or simply used an AI generator to mimic his voice. Neither artist has yet confirmed or denied either way, although Drake did share the song on his Instagram story, hinting at his approval of the track.
The song is widely held to be at least partially a parody. Yet the bigger story is the use of AI itself, and the fact that no one can tell the difference.
The internet is divided, with audiences and artists alike up in arms over how seriously to take the song, whether Drake approved, and whether it is AI or not. There have been plenty of AI uses aside from Drake's own history that have blurred the lines, from Johnny Cash’s voice covering a Taylor Swift song on YouTube to the Beatles’ official posthumous release of ‘Now and Then’. Yet in these cases, creators themselves have publicly declared the use of AI in the tracks. With ‘Wah Gwan Delilah’, however, we simply do not know – and there is no way of telling.
When investigating the story, Billboard contacted two companies that specialise in AI audio detection to get to the bottom of the track’s origins. Despite the fact that there should be plenty of data on which to run the analysis, given the plethora of confirmed songs both from Real Drake and Fake Drake, neither company was able to provide a definitive answer beyond the fact that some AI may have been used.
Featured Report
Visionary audio Unlocking the power of video in podcasting
YouTube may be the only viable platform for long-form video podcasts, but that does not mean audio-first podcast platforms should abandon video. Instead, podcast platforms should leverage video both as...
Find out more…So we are now in an environment where an artist who has been burned by AI seems to have embraced it, a new song featuring his voice has dropped and no one can tell if it is real or not, and the companies that offer any hope of clarity in an increasingly murky digital environment really cannot seem to do much to help.
The real threat of AI comes from two fronts: one is that the sheer volume of tracks (or digital content overall, because the impacts here are mirrored across entertainment) dilutes the value of music in general. The other is most commonly referred to as ‘deep faking’, but it really is a broader issue of unreliability. In a digital world dominated by AI generation, there is simply no way (yet, if it is possible at all) to verify what is real and what is not – and the AI generators have been released too early for there to be any serious built-in safeguards against faulty information, copyright infringement, or general malpractice. Artists must either embrace the chaos and see what happens, stay offline altogether to retain control over their voice and image, or engage in an ongoing whack-a-mole of legal battles to seek clarity and enforce ownership of their rights. In the meantime, audiences are confused and increasingly distrustful. AI generators are still not generally profitable, and the digital world is overrun by its use despite all the flaws that are inherent. These issues are now inescapable, as AI continues to be trained on web-scraped data, which is increasingly populated by AI-generated content itself.
Drake’s image is big enough that he can afford to take the punches and experiment in the new AI-powered world without it affecting his clout (the quality of ‘Wah Gwan Delilah’ as an actual song notwithstanding; it may be deemed cringe, but 'BBL Drizzy' is a verified banger). Indeed, he is benefitting from the online debate in a classic example of ‘no press is bad press’. Yet for less established artists, this is a far more complicated environment to navigate – and far more threatening. Creators like ghostwriter and potentially Snowd4y can use the voices of famous artists to boost their own profiles, but those who want to stand on their own will struggle to compete with the growing artificial presence of the already-biggest players. And audiences may start to disengage entirely, not sure whether they can trust the reality of anything they see on streaming or social – any new drop could be entirely faked. Authenticity is of increasing value in such an environment, but current AI innovation seems to be set to corrode it. Live streams of artists playing and mixing will likely become more important in reassuring audiences of the reality of releases (at least until AI generators become better at video). Analogue consumption may start to increase in response as well as interest in live shows – but cost issues can stand in the way.
AI is an inflection point the industry was not ready for. Legal proceedings have a critical role to play, but they take time – and in the meantime, Pandora’s box is wide open. Artists must adapt to survive. They can embrace AI completely, without certainty of what will work and what comes next, but knowing it has a role to play – or avoid it entirely, on the premise that the digital boom is receding. The future will be a mix of both, and what creators choose to do with it will inform the status quo that solidifies, much as LimeWire piracy opened the gates to streaming.
And in the midst, we will continue to live in a digital Shroedinger’s AI box, where no one is quite sure whether Drake is really Drake.
There is a comment on this post, add your opinion.