Rage and the machine: Fan generates six albums of RATM-style jams using AI – and gets real musicians to record them
The AI Kittens project was inspired by Tom Morello’s 15-minute songwriting challenge
A Rage Against The Machine fan named Marat Nigametzianov has created a machine learning network and set it to do god’s work: generating 1,500 RATM-style ideas a month.
Nigametzianov was reportedly inspired to take the challenge back in January 2020, when he attended Tom Morello’s Guitar Revolution event in LA. Morello led a workshop in which five people were chosen to write a song in 15 minutes, with the help of mentors like Wayne Kramer, John 5 and Vernon Reid.
“When Tom gave his mentee his Black Spartacus [guitar] – [the] dude almost fainted,” says Nigametzianov in his post on Reddit. “This event was to inspire people to make music, [to] write and record their songs. I’m not a musician, but I was very well inspired.”
Some months later Nigametzianov heard about the release of the Open AI’s Jukebox – ‘a generative model for music’, which provides an open source basis for generating music via machine learning or ‘neural network’.
He soon set to work building servers and testing the capabilities of a burgeoning Rage machine: feeding it selected Morello and RATM cuts and steering it via the code to generate Morello-esque riffs and other samples.
“I started to generate one minute samples [at a] speed of 1500 samples per month,” Nigametzianov says. “[In total now I have] generated around 12k samples to this date. I personally listened to all this stuff and picked samples with song ideas.”
Listen to Riffs, Bass, Meow, Vol. 1 on streaming platforms
Get The Pick Newsletter
All the latest guitar news, interviews, lessons, reviews, deals and more, direct to your inbox!
Nigametzianov then put out the call for willing musicians to participate and recruited three ‘teams’ of musicians and engineers – including guitarists Anton Dokuchaev, Rostislav Chaban and Mick Lisov.
The best of the samples were then grouped up and sent out to the musicians, who recorded six album’s worth of the material.
You can hear the results on streaming platforms under the band name AI Kittens.
We’re not sure RATM need to worry about AI Kittens threatening future bookings, but there is something weird and wonderful about listening to this material and considering the melding of man and machine learning that created it.
We wonder what Morello will make of it. At the very least, we figure he’ll have to appreciate the phenomenal, none-more-internet album cover of Music Samples in Style of RATM. Vol. 1...
Listen to Music Samples in Style of RATM. Vol. 1 on streaming platforms
“All this was done for my own fun and for the fun of RATM fans all over the world,” says Nigametzianov. “It is not intended to offend anyone. Just listen and have fun if you like it.”
If you’re interested in learning more about the project, or hearing some of the original samples generated by the neural network, check out Nigametzianov’s full post on Reddit.
Some quoted material in this story has been adjusted for ease of readability.
Thank you for reading 5 articles this month**
Join now for unlimited access
US pricing $3.99 per month or $39.00 per year
UK pricing £2.99 per month or £29.00 per year
Europe pricing €3.49 per month or €34.00 per year
*Read 5 free articles per month without a subscription
Matt is Features Editor for GuitarWorld.com. Before that he spent 10 years as a freelance music journalist, interviewing artists for the likes of Total Guitar, Guitarist, Guitar World, MusicRadar, NME.com, DJ Mag and Electronic Sound. In 2020, he launched CreativeMoney.co.uk, which aims to share the ideas that make creative lifestyles more sustainable. He plays guitar, but should not be allowed near your delay pedals.
“You can only imagine the effect this had on the young Keith Richards and Eric Clapton”: 9 must-hear albums that fueled the British blues guitar boom
“We’ve made something really unique and special”: Thin Lizzy to release first new record in over 40 years – featuring brand new guitar parts from founding member Eric Bell and unheard Phil Lynott vocals