Music
Artificial aesthetics: Meet the UK musicians experimenting with AI
AI is both awe-inspiring and unnerving. Will the technology ever replace human musicians?
We probably associate artificial intelligence most with sterile algorithms, post-work utopias or paranoia about future wipeouts of the workforce – or that Steven Spielberg movie from 2001. Another area where it’s developing is music.
Cutting edge producers in the UK are unleashing the strange and beautiful potential of collaborating with AI creatively. In terms of genre, much of the best work coming out of the UK has an ambient or IDM bent, and is by turns powerful, affecting and strangely unsettling. These AI music projects are taking us into an uncanny valley of sparse sonics and emotional triggers.
It’s too soon to claim there’s anything like an AI “scene” in UK music yet, but there is a wave of curious artists who are experimenting with the technology. Read on to find out what’s driving three of the main pioneers and what their visions are for the future of AI’s role in music.
Essentially music is a reflection of humanity. It’s not going to interest us if it’s completely devoid of that.
Ash Koosha / Yona
Ninja Tune’s Ashkan Kooshanejad is a modern-day polymath who can’t be pinned down to one musical style or identity. The albums he has released under the moniker Ash Koosha are feats of aggressively experimental creativity, veering between inscrutable mayhem and glacial beauty, like whole ecosystems blossoming and decaying in a timelapse. Growing up between Tehran and Frankfurt and now based in the UK, his career has been punctuated by time spent under the radar, toiling away at his experiments. But he has nonetheless been prolific in producing four albums since 2015 (with two of those arriving in 2018 alone) and has a highly sophisticated understanding of how computerised technologies can supplement human inspiration.
Koosha’s record Return 0 (whose name comes from the C++ coding language) was the culmination of his research on AI, when he realised that he could draw on it to collaborate meaningfully. It went well, but he’s now moved on to new work. “With that the Ash Koosha brand is kind of closed for now,” he says. “I've moved on to this new project that I developed last year called Yona”. Yona is a kind of living, breathing digital avatar, but AI is only one technology used to bring her to life.
“It is only a few subsets used in a variety of software and engines that we use to make this happen”, he explains. “It's from audio engines to NLP [Natural Language Processing], markov chains, a lot of different things come together. It's a sort of combination of software that generates this person. Now the difference with what I've done before is it's personified through this character that will eventually reveal its backstory.”
While Kooshanejad’s work focuses on turning Yona into an independent performer, he’s clear that this is a human-led creative process. His music isn’t science fiction. “People want to hear something that doesn't sound human and like it's a robot. But I don't believe in robot cognition at all. Maybe give it 50 years, but I still believe in machines as tools and supplements to our processes… The brain of a machine is not even close to cognition. It's a replicated set of computational work at play but it's not near understanding or intention. Intent is the most important part in making or composing music.”
That may be so, but when Yona performs her first show at Rewire festival in The Hague in March, and releases a record or at least a few singles, it might be hard for the listener to draw the line between where artificial intelligence begins and the human input ends.
I take the promise, threat and particularly the hype surrounding AI all very seriously.
Kode9
Kode9 performs at Red Bull Music Festival in Tokyo, 2017
© Suguru Saito / Red Bull Content Pool
Steve Goodman is a Scottish, London-based intellectual who runs the Hyperdub label and, for his own music, uses jagged electronic soundscapes to explore the dystopian potential of technology.
Goodman is interested in the feeling of dread caused by the concept of AI. “The promise of ugly, strange and inhuman musical aesthetics that humans wouldn’t even bother to dream up is very exciting,” he says. “I take the promise, threat and particularly the hype surrounding AI all very seriously.”
That threat is military and political. In his 2010 book Sonic Warfare Goodman examined the use of sound to influence and control populations, covering everything from sonic booms over the Gaza Strip to high-pitched frequencies in shopping malls. While those phenomena don’t involve AI, his new work will marry the two, and not for the first time: read his dystopian manga about AI overlords and sound control here.
One of Goodman’s upcoming project has been inspired by “the recent news stories regarding sonic attacks on US embassies in Havana, Cuba, and Guangzhou in China”. In response, he’s been working on the connections between AI and voice. “We have been developing automated systems that are capable of generating statements and tweeting in a number of voices, and sonifying these using text to speech software”.
That’s all part of the AUDiNT project that Goodman works for with audio-visual artist Dr. Toby Heys. A self-styled research unit, it riffs on 20th Century military experiments where researchers explored how sound can be used to influence people. “For the past 60 years, AUDiNT has been conducting research, rituals, and experiments into the opening of the 3rd ear, a dimension that materialises when sound, ultrasound, and infrasound are simultaneously deployed,” its website reads.
One of Kode9’s future aims is to channel those mysterious signals into his music through an ongoing project called 3rd Ear Cat, which features animation by Kim Laughton. Modelled on a Japanese “waving cat” figurine (known as Maneki Neko), the 3rd Ear Cat is an AI character that channels infrasonic and ultrasonic signals into its “meat puppet”, Kode9. In the manga mentioned earlier, it’s imagined as a God worshipped by earth’s AI overlords. But it may end up being a focal point for Kode9 material in the real world one day. That “might take a couple of years and more developments in machine learning to properly come to fruition,” Goodman teases.
If you’re hungry for some Kode9 headfuckery in the meantime, look out for the Barbican’s AI-focused exhibition, More Than Human. That’ll run in London from 16 May til 26 August. His contribution, a piece titled IT, is inspired by the old Jewish myth of the golem in which a creature is brought to life out of clay. In the folklore, the creature is activated by the power of language and goes on to run amok. It’s a running motif in sci-fi stories of AI that have influenced Goodman, and he wants to use the piece to explore the religious connotations of singularity theory, an undervalued element of what is often seen as a purely scientific topic. All in all, it’s bound to be another brain-melter from one of the UK’s foremost thinkers on the topic.
Sevenism
Sevenism is a Nottingham-based producer whose leftfield ambient output has stirred up a cult following. He is beyond prolific, utilising AI to create at a speed and ferocity no human realistically could. “At the moment I’m doing an album every week or two,” he smiles. His records have names that mirror the sort of wild poeticism familiar to fake text generators: each raindrop is a sphere (that scatters light over the sky) is one example. He’s a poster boy for running with the technological and poetic possibilities of these new softwares in an irreverent, unapologetic way.
As a child, he would dream of hooking his brain up to a computer and somehow feeding the music that he was creating in his head directly into it. In the late ‘90s he did a bit of AI and psychology at university, but switched to straight psychology when he realised how much he preferred the artistic and philosophical possibilities to the reality of programming. That led to a period of creative absence, but recent leaps forward in technology have allowed him to re-engage with his interests. Google’s Magenta project, for example, has created ways for machine learning to complement the creative process.
There are lots of different AI technologies that have been created as part of Magenta. There’s NSynth, a neural network which draws on 300,000 sample sounds to generate new sounds, or Piano Genie, which converts inputs from eight buttons into the sound of an 88-key piano. According to Magenta, Piano Genie is “in some ways reminiscent of video games such as Rock Band and Guitar Hero… with the crucial difference that users can freely improvise.”
“There are really great tools that get you instant results really,” he says. “There’s stuff you can plug into music programmes, just fire up and get going.” Rather than sitting on GarageBand or Logic painstakingly laying down a musical idea he has had, Sevenism uses these AI technologies to set the process in motion. “The inspiration can come from the programme itself. You don’t necessarily have to input anything. It can start off and the inspiration can come from the instrument I suppose.”
If AI makes the process of making music easier, he suggests, that doesn’t cheapen the end result, it just expands the possibilities of what can be produced. “The way that the AI’s been implemented in traditional software means that you can produce really great stuff a lot more easily than in the past,” he says. “I think people are stuck in that idea that you have to kind of work, work, work, on a track for months in order to reach perfection, but I don’t think that — I think it’s a lot easier to make something really good now, almost improvisationally”.
This can result in a sudden flash of interest in a particular track online, with people hearing something Sevenism himself didn’t even notice. It’s a nice surprise, he says, but he’s not particularly interested in looking back. So does he think the future of music is generative, devoid of humans altogether, producing itself according to an algorithm in a linear line in the way that some have speculated could become the case? “I think that’s a gimmick. I do not think people are going to be interested in that.” For Sevenism, there’ll always be an element of human control in the music, down to the coders who first set the algorithm into place. “Essentially music is a reflection of humanity”, he says. “It’s not going to interest us if it’s completely devoid of humanity.”
Like Red Bull Music on Facebook and follow us on Twitter at @RedBullMusicUK