DigitalFruit is an interview series from Adafruit showcasing some of our favorite digital fine artists from around the world. As we begin this new decade with its rapidly changing landscape, we must envision our path through a different lens. Over the next few weeks we’ll feature many innovative perspectives and techniques that will inspire our maker community to construct a bold creative frontier. The only way is forward.
1. Where are you based?
My name is Amir Zhussupov. I was born and live in Almaty, Kazakhstan.
2. Tell us about your background?
I graduated from a local university with a major in Corporate Finance, but that’s a boring story and I’d much rather summarize my creative path.
When I was in 10th grade, I was really passionate about music, so I started learning about composition. I was a huge fan of electronic music, and especially drum and bass at the time, so I started with producing liquid funk. But I didn’t stop there, I was experimenting with other genres, trying to find something that I might call my own style. After some time, I didn’t feel like it was where I saw myself in the future, so I stopped, although I was still DJing for my friends every now and then. Some of these skills are quite useful for me up to this day. After music, I had some thoughts about game development, I even made a couple of simple games from scratch, but being an indie-developer seemed to me an incredibly unstable and stressful lifestyle, so I did not fully commit to that.
I still had a desire to create, which I have to fulfill one way or another, so luckily during quarantine I stumbled upon a tutorial on generative art in Unity. It got me very excited, I remember spending the next week purely watching lectures and reading articles about it. I also felt that if I started doing generative art, I might as well share it.
At first, I was planning to focus on creating art in Unity or TouchDesigner, but somewhere in the second half of April I learned about Generative Adversarial Networks (GANs), particularly those that generate images, and eventually I figured out how they work and how they can be used for artistic purposes.
At its core, GAN is basically two neural networks called generator and discriminator, where the first one generates data and the latter is trying to distinguish between real input and generated output. They constantly learn from each other, and over time, if the training goes well, a generator starts producing results that are almost indistinguishable from the training dataset. That may not always be the case due to insufficient training time, faulty datasets and other things.
3. What inspires your work?
First and foremost, the idea of GAN training. It starts with “feeding” the dataset into a model, training it, and finally generating output. For me it’s not dissimilar to how we, humans, learn by absorbing information, which shapes our thoughts and ideas, thus ultimately forming our own self-expression.
One of the other reasons that I am so drawn to AI art is curiosity. Like, what would an AI create if it was trained on the works of H. R. Giger? Or album covers? These are not perfect datasets by any means, but it’s enough to get an interesting output from the model.
Another huge inspiration for me is, of course, the work of other people, with whom I share the same hashtag space on Instagram. Also, since everything I do is digital, I feel like there is a huge influence from the video game world. For example, when I was a teen, I really liked a game called Audiosurf, which conceptually reminds me of a lot of audio reactive/audiovisual stuff that new media and generative artists create. And of course, music, which I’m using to set up a certain mood for myself and my videos.
Speaking of them, people seem to particularly enjoy my latent space animations, and to understand how they are made you’ll have to imagine what a latent space is. In simplest terms, think of it as a space where, after the model is trained, all of the generated data (images, in my case) exist simultaneously. Thus, a latent space walk is basically an animation where each frame is a different image from that space.
4. What are you currently working on?
In the future, after the whole COVID-19 situation, I’d like to explore the possibilities of doing live performances and installations, not necessarily AI related, but something fun – so I’m learning TouchDesigner and Houdini. I also have a long-term AI collaboration going on, but I can’t share any info at the moment.
5. Describe your process and what tools you like to use.
In addition to the software that I mentioned above, I’m using Premiere Pro for all of my video editing. It’s very intuitive, and covers all of my needs. For audio related tasks – Ableton Live. I’ve been using it for over 7 years and its workflow is pure joy. For fast dataset preprocessing, cropping for example, I’m using a little program called BatchPhoto, and if I need something more specific, I may use droplets in Photoshop. For my coding purposes, it’s usually Visual Studio Code.
6. What does your workplace or studio look like? Do you work in silence or listen to music while you work?
I work at home, usually in my bedroom. You don’t need a studio or any special hardware apart from the PC itself to do stuff at my level. When I create, let’s say, looping latent space walk videos, I try to accompany them with music, so I listen to the same 4-8-16 bars over and over again making sure it’s seamlessly looping and fitting the mood.
While I wait for model training or rendering, I usually just watch tutorials. I have around 30-40 tabs in Chrome set up. When I finish some, I find new ones to replace them very quickly. It’s a never-ending cycle ever since I became interested in generative art. So, I’m doomed with Chrome eating most of my RAM at all times.
7. How has technology shaped your creative vision?
Well, A LOT, haha. I never had a knack for painting, so I never saw myself as a traditional artist, never even bothered with that. Thus, knowing that you may create something without any drawing skills was pretty encouraging to me. Technology, i.e. software, has definitely made any type of art a lot more accessible for people of all kinds of backgrounds.
8. Any tips for someone interested in getting started in the digital art form?
If you want to just have some fun with generating AI images online, check ArtBreeder. There you can create latent space walks similar in nature to mine, as well as “breed” images together and so on.
For a more serious approach, which is an actual training of a model on your PC, you’ll need to pick and download a GAN that works with images, a good GPU to train it (preferably a bunch of them), or a thick wallet if you’d like to use cloud services for training. So actually, you’ll need a thick wallet in both cases 🙂
9. Where do you see generative and digital art heading in the future?
Generally, I hope that both artists working in that sphere and the techniques they use will be getting more and more recognition from the traditional art community. I feel like the true beauty of generative/digital/new media art unfolds not on the screen of your device, but in the form of performance, installation, projection mapping.
Talking about AI art – I remember how in 2018 a GAN painting was sold on an auction for ~$400,000, and there were a lot of critics who did not take it seriously. It’s been two years, and apart from some technological advances in that sphere it’s nowhere near mainstream yet. Does it have to be, though?
DigitalFruit is curated by Adafruit lead photographer- Andrew Tingle