AI turns The Spice Girls' "Wannabe" into a pretty solid Nine Inch Nails song
Music Features Nine Inch Nails
Artificial intelligence programs, once pretty clunky and easy to mock, are getting really, really good at their jobs lately… like, “straight-up just rewriting our concepts of reality and making us question our meat puppet mentalities in this ‘first world’ country of ours” good. It’s a terrifying prospect for most of us, but after over thirty years of industrial, dystopian goth noise rock, that’s exactly the kind of thing that sounds right up Nine Inch Nails’ alley.
Well, Trent Reznor, let’s see how much you like it when an AI program gets fed the sonic antithesis of your life’s work and approximates your own art, shall we? Behold: The Spice Girls’ “Wannabe” in the AI-style of a Nine Inch Nails song.
Aw hell… that’s not half-bad if we’re being honest. Created by YouTuber JohnDopamine5 using the OpenAI Jukebox system (a program that has churned out equally creepy, catchy work in the past), this “Wannabe” by way of The Downward Spiral sounds pretty damn close to its muse, complete with Reznor-esque vocal growling between off-kilter synth tones and electronic drum beats. It’s only about 30 seconds worth of AI-generated NIN, but it’s enough to make us curious to hear an entire track, or even—dare we dream—an entire album.
To be fair, if there ever was a band to release an LP co-written by our future AI overlords, it would almost definitely be Nine Inch Nails.
[via Digg]
9 Comments
I thought the title said Al instead of AI and figured it was a Weird Al Yankovic story. Was disappointed.
Simpsons did it! Er… I mean someone else did it. Years ago in fact.
The vocals are the creepiest part to my ears. Really sounds like it could be Reznor himself.
It’s unclear to me, from the article and video, what is being generated. The link to other output is a video where it’s made very clear that the composition is AI, not the audio.
As I understand it this is both composition and audio. It’s what we call a “deepfake” but for music.
OpenAI Jukebox was trained with at least the one Spice Girls song and was asked to continue the composition. At that point the new “composition” is data that’s sort of like sheet music but in some weird AI format.
If this composition was fed back through the Spice Girls audio training set you’d get something very similar to the original song. Instead of doing that, they fed the composition back data through an audio training set generated from Nine Inch Nails albums.
Trents been known to do covers, and very well, but he takes himself to seriously to do anything like this
Nope
That. Was. Terrible.
How are we defining ‘AI’ these days? Twenty years ago it was anything that could pass as human. These days it’s ‘any piece of software that makes something shareable on the internet’. Faster processing power means you can get a piece of software to analyze a dataset faster and perform a required function as a result of that dataset faster than it could twenty years ago, but that doesn’t make it AI. This piece of software just takes lyrics and rhythm and re-synthesizes them to match a different key and pitch, using mashed up audio samples. It’s fun, yes, but it isn’t AI, it’s just a few spreadsheets and a lot of processing. It would be AI if I just typed in ‘set the lyrics of a Spice Girls song to a Nine Inch Nails type sound’ and it did all the research, the thinking, the trial and error and the creative work itself.