Robot Boat Party: YACHT’s AI Experiment
Thus far in history, music has typically been written by humans. YACHT bucks that trend quite well with Chain Tripping, a project written mostly by machine learning algorithms. However, the success of the album shows that there are many problems that the music industry will face as technology grows.
Written by Adithya Srinivas
When dance pop outfit YACHT — producer Jona Bechtolt, vocalist Claire Evans, and instrumentalist Rob Kieswetter — finally decided to record their next studio album, they also decided they didn’t want to do much of their own writing. No, they didn’t hire ghost writers, make a cover album, or just steal some other band’s hard drives. They decided they would let robots write it.
YACHT partnered with leading artificial intelligence and machine learning experts from Google and others formerly employed by the Obama administration to create a program that would write most of an album for them. The algorithm that they ended up with was fed MIDI inputs from all 82 songs in the band’s commercial discography, along with 2.2 million lyrics from artists that they considered influences on their own projects. The code then went to work dissecting YACHT’s sound, finding patterns and tendencies that defined their music, and every so often, it would take what it had learned and spit out its best attempt at a YACHT composition, which the band would record.
Judging by the creative process alone, you might expect the album to be a collection of tracks that are simply the least common denominator of YACHT’s songs, or not sound like them at all. You might expect the lyrics to be completely out of Claire Evans’s character or entirely incoherent. You might expect the compositions to be boring and formulaic — after all, how musically creative can a computer be? None of those impressions about an album written by code are unfounded, and yet, none of them end up being true.
The result, Chain Tripping, is an almost scary extension of YACHT’s sound, as it stays true to their auditory aesthetics while still being a unique progression from previous albums — and they didn’t even write 90% of it. The lead-off song “(Downtown) Dancing” immediately roots itself in Bechtolt’s signature ultra-groovy bass loops and Evan’s spaced out vocal delivery. But there’s a very stripped-back feel to the track compared to older YACHT songs, a style that carries through the entire album. Apart from adding some tasteful cowbell, tambourines, and the racing, spooky synth hits during the chorus, there’s really nothing else on the track, yet it remains a full and danceable song. The next track, “DEATH,” happens to be the slowest song the band has ever released, yet it still maintains enough momentum to remain a fun beat. “Stab a cop” as a hook is a tad bit unsettling (especially coming from a bot); however, the darker lyrical themes throughout the song are almost lost in the jovial bass line, acoustic guitar riffs, and airy synth piano.
Later in the track listing, we’re met with “Sad Money”and “SCATTERHEAD”, both of which could have just as easily been on YACHT’s previous album, 2015’s I Thought the Future Would Be Cooler. This isn’t a bad thing — rather, it’s a testament to how well the program was able to replicate the band’s more complex dance and electronic arrangements — and the more grimy and simple composition style still sets these songs apart.
Of course, it’s not all perfect. The songs “Hey Hey” and “Loud Light” definitely feel like they were written by a computer, both featuring clunky and unnatural vocal passages that Evans executes well, but that still come off as awkward. The closing track “Little Instant” is actually a quite underwhelming finish, being the least melodic or fun part of a great dance album. Regardless, Chain Tripping is easily the most surprising in YACHT’s discography.
The success of this experiment will most likely catch the eye of other producers and artists, but more widespread use of machine learning could open a Pandora’s box of unforeseen issues. YACHT obviously commissioned their own algorithm, but what if an artist doesn’t ask for one? What if, for example, Michael Jackson’s estate decides to commission an algorithm of their own? They could find a talented vocalist with a similar voice to record the algorithm’s compositions (*cough* The Weeknd *cough*), hire experienced sound engineers to doctor the vocals to match Jackson’s, and release a new “Michael Jackson” album. Combine this with the increasingly popular hologram concerts in “world tours,” and you could almost resurrect an artist who has passed.
The ethics of this are dubious at best, but no one has ever accused the music industry of having a strong moral compass. Corporate executives and managers will surely find a way to use this technology for their own financial gain, whether that means exploiting artists who have passed or stealing sounds of talents outside of their own labels. As computing advances to stranger frontiers, these dangers only become more prevalent in the music industry.
However, Chain Tripping has shown that while there is a lot of promise in artificial intelligence, it’s far from a perfect technology. As a result, we’re probably a long way from “Black Mirror”-esque musical dystopia, but considering the future consequences of an emerging technology is an important process in determining whether it should be widely used at all. For now, machines have given us an entertaining and replayable electronic dance-pop album in Chain Tripping, so go check it out before the robot uprising.