“These are important visual moments”: artist Robbie Barrat pushes, tests and breaks AI in his works
With a collaboration with Acne under his belt, Robbie talks to us about his plans for the future, which include attending art school for the first time.
If there’s one word to describe new media artist Robbie Barrat, it would be ardent. Currently living in Nantes, France, the West Virginia-born creative has had a distinguished career for someone who’s only 19. Straight out of high school, he worked in a research lab in California before moving to New York where he practiced as an artist, he’s hosted an exhibition in collaboration with artist Ronan Barrot and, most recently, collaborated with Acne Studios on its AW20 collection – all projects which have employed AI or machine learning in some form.
While it’s a bafflingly impressive start to his career from the outside, during our conversation with him over the phone, it becomes clear that it’s simply a product of pure absorption in a field – and a whole load of talent as well.
Growing up, Robbie was naturally interested in robots and was always on the computer as a child. He tells us how he “sort of always knew he wanted to be an artist” but it wasn’t until he started working at NVIDIA in the Bay Area in California that the decision was solidified. “It sucked actually, to be honest,” he recalls. “It was awful. Because I was 17 and I didn’t really have any friends my age. All my friends were, you know, 40-year-old guys.” He had first come into contact with AI while still at high school, when he saw some images from an early paper on GANs (generative adversarial networks). “It was a purely technical demonstration but I was really obsessed with how these generated images would have stuff that looked like text in them but it wasn't readable text,” he says. “And then shortly after I started generating images with paintings.”
Often, when explaining his work for the first time, Robbie will refer to a series of these paintings which he completed in early 2018. Starting with landscapes, he attempted to use a GAN to generate paintings and was able to get fairly realistic results; “but realistic results are very boring after about ten minutes,” he says. It was the discovery that the most compelling things happened when the GAN misinterpreted the information, which really shaped his practice today. He began to experiment with creating nudes where the network was able to “correctly learn ‘rules’ associated with small and local features of paintings (breasts, folds of fat, etc) – but failed to learn rules concerning the overall structure of the portraits (two arms, two legs, one head, proportions, etc).”
These techniques were employed last year on Robbie’s biggest project to date, a collaboration with Acne Studios on its AW20 collections which launched on 19 January. When asked how the hell this project came about, Robbie explains that in June of the previous year, Acne reached out to him. The team had seen his previous work using Balenciaga data, in which he created entire collections by training an AI on a data set full of Balenciaga’s previous collections, and essentially, Acne wanted to do something similar – but for real, and not just as an experiment. “So, I went to Sweden a few times and worked with them,” Robbie tells us. “I produced images for them – it was really a collaboration with their designers where they would look at the images I was producing and then try to interpret these into outfits.”
1 of 6
Balenciaga Neural Network
Robbie trained a neural network on the previous four seasons of Acne’s collections in order to create new designs, and he also produced imagery which was directly printed onto garments. Alongside this, he developed tools for the designers themselves to use, where they could click on an area of an outfit and alter or “correct” it somehow. The result is a series of garments which remix the notion of the collar, or the hemline, for example, simply because their concept was produced by an AI with no concept of what they should look like. “I’m really happy with the end result,” he says on this point, “but I’m a bit sad that it wasn’t a bit more bizarre.”
A lot of the imagery Robbie produced was off-kilter; sketches of clothing in which you could recognise it was a garment, but it didn’t have the characteristics of a jacket or a coat at that stage, “it was just a big lump of clothes.” Initially, this concept had really inspired Robbie – “I really wanted to do something with clothes that you couldn’t wear but I was quickly pulled aside by a marketing person and told that wasn’t going to be possible,” he jokes. With further experimentation, though, it proved just as interesting, if not more interesting, to maximise the elements that the network misinterpreted and it was this approach that led to some of the more “visually fresh” components of the project. Ultimately, the collection is an interesting experiment into the application of AI into a design process, but it also raises an important misconception about AI.
Often, when we talk about artificial intelligence, machine learning or neural networks, a certain agency is granted to the tool. But Robbie is quick to point out that it is just that: a tool. Some press surrounding the show, for example, has described the collection as one concerned with neural networks and their application within the creative process and “that’s sort of what it’s about,” Robbie says, but, really, “it was a collaboration with me as an artist, I was doing artistic work for them, I wasn’t just the ‘tool guy’.” While AI obviously presents a lot of potential for a brand, say, working within the creative industry, what presents more potential is for that brand to collaborate with an artist who works with AI. “Acne could work with someone else – a person or company or another AI artist – and get garbage, or at least something completely different,” Robbie says. “There’s nothing about the tool which is particularly interesting – there has to be a collaboration with an artist first.”
“There’s nothing about the tool which is particularly interesting – there has to be a collaboration with an artist first.”Robbie Barrat
While the Acne collaboration has certainly introduced Robbie to an audience who might not have previously discovered his work, the artist’s name appeared in a fair few articles well over a year before, in October 2018, when tech startup Obvious sold “the first AI artwork” at Christie’s, and it later turned out the group had stolen the code from Robbie. Initially estimated to fetch between $7,000 and $10,000, the artwork titled Portrait of Edmond Belamy eventually took $432,500. Obvious admitted to using Robbie’s code but how much of it is ambiguous. The Verge reported that “experts say the amount was probably substantial,” but whether Robbie could claim ownership of the work caused heated discussions as he shared the original project under an open-source license.
When asked for his perspective on the situation now, Robbie says: “The group that sold that work was not an artist group, they were a marketing group. I know that sounds like an attack but if you look at early articles that they've written, they refer to themselves as a startup. Honestly, they really did take advantage of the project that I put up, even though it was open source.” The outputs, he continues, were just so similar to his that even if the issue of the ownership of AI or source code were removed, it was “a ripoff of my very early, very bad work”. So not only did he lose out on a “tonne of money,” but it exposed to the world “a lot of the bad work I did when I was 17,” he adds in jest. “Plus, they attached this horrible narrative to it about how the computer made the work so, now, every time I speak I have to take 15 mins at the beginning to explain: ‘By the way, that's not really what happens.’ They’re just saying that because I guess it sells for more if you say that it’s the first robot-authored painting.”
It’s a complex area, though, because that code was out there. But it was out there under the pretence that most members of the open source community understand the unwritten rules. “I don’t have a good answer for that,” Robbie says, when asked about how you can protect your intellectual property while also contributing to this community. But, if you’re ever in doubt, “just ask the author of the code if it’s OK to use it to sell something that you're producing with it,” he adds. There are plenty of people using his code to produce work – and they’re selling their work – the big difference with the Obvious case, he concludes, was “the visual similarity on top of everything else,” and that “nothing that they did was motivated by artistic sense.”
However, he’s not deterred from sharing his work, including its code, online: “I really hope people don’t look at this and view it as a failure of open source or a reason to not make their projects open source. There are still a few projects that I really need to clean up the code of and, when I do, I really want to upload them and make them open source. I mean, I’m going to be a bit more careful about it – I’m only going to upload tools essentially.”
It’s at this point that we should point out that Robbie is a self-taught artist and coder, and that he only recently took the decision to go and get a bachelor’s. When we catch up, he’s just come out of a class at École des Beaux-Arts de Nantes Saint-Nazaire, a university he’s been attending for the past five months. He’s undertaking a one-year-long preparation course for foreign students who want to go to art school in France – it’s kind of like an art foundation where they also teach you French. (“How’s your French?” I ask – “It’s shit!”)
With his skillset, you’d forgive Robbie for pursuing a degree somewhere like ECAL or RISD but, instead, he’s after a formal education in fine art. “I think you can only get so far by ignoring [traditional media] and only working on digital,” he says, providing an example from his time working with Ronan Barrot last year: “Part of his process when painting was to ‘correct’ part of his own paintings by covering up part of them with bright orange paint and then filling that area back in.” This heavily influenced Robbie’s own process and he went on to create an entire series titled Corrections where he used AI to do the same thing. “I feel like if I were really exposed to traditional media over several years, it would be like that x 100! There are so many interesting things that I can try out, borrowing ideas from sculpture or other disciplines,” he continues.
“You can only get so far by ignoring [traditional media] and only working on digital.”Robbie Barrat
So why would some already establishing themselves in the industry and achieving the things many grads crave – collaborating with brands and hosting exhibitions with international artists – choose to go and study? “I feel very lacking in traditional art education and traditional mediums,” Robbie says. “And also, admittedly, it was what I was telling you earlier about how working in industry was pretty depressing. I mean, I just want to be around other artists my age, I don’t want to be working in some sort of void.”
For now, his attention is on old 3D graphics, arcade games like Big Buck Hunter or gardening and landscaping softwares. “I’m exploring glitches. Like if you accidentally put the camera, the viewpoint inside of a shape, and then you can see some edges of the shape. I think that there’s some really interesting visual artefacts that happen in stuff like Big Buck Hunter but nobody really pays attention, because of the fact that they happened in Big Buck Hunter,” he explains. It’s a process which sees him using these softwares and thinking about them laterally. One way he’s been doing this is by recreating existing paintings, and then looking at it from a different angle, or from underground, for example. “I feel like these are really important visual moments,” Robbie remarks. “It really reminds me of Cubism.”
There are also whispers of another collaboration with a different fashion brand, but really he wants to dedicate his time to developing as an artist. “I feel somewhat frustrated because a lot of my work has been about AI as an artists’ tool and that’s cool, but it’s not what I want to do until I die,” he says. “I have a lot of what I think are really good ideas which use AI in them but they’re not just about AI.”
Ultimately, what I love about Robbie’s work is how he just seems to get AI and new media in a broader sense. He understands its capabilities and potential from an artistic standpoint but he doesn’t hold it on a pedestal or claim that a work is good just because AI has been involved in its making. What he does, instead, is see it as a tool which can be pushed, tested, experimented upon and broken in order to achieve interesting results. And it’s not just any tool, but one that has so many possibilities by its very nature, and which currently feels so untapped within the creative world, like we’re only seen the tip of the iceberg of what will one day be looked back upon as an entire movement. To borrow a comparison made by Robbie himself, “it’s similar to the work of Sol Lewitt, where he left the rule card to others – I feel like I’m passing the neural network the rule card.” And with Robbie writing the rules, it’s safe to say we can expect exciting things.
Spotify Design is proud to sponsor It’s Nice That’s Ones to Watch 2020, championing a diverse and inclusive creative community. We’re a team of cross-disciplinary creatives who love to design experiences that make meaningful connections between listeners and artists. We’re committed to designing for tomorrow and to supporting future-facing creatives who are out to make an impact on the world.
About the Author
Ruby joined the It’s Nice That team as an editorial assistant in September 2017 after graduating from the Graphic Communication Design course at Central Saint Martins. In April 2018, she became a staff writer and in August 2019, she was made associate editor.
Ones to Watch 2020
Read more profiles of this year’s Ones to Watch