Moving between Computational Creativity and Creative AI: the Art of Matias Nordin
Explorations of AI Art — Episode 12
“AI is today already a creative tool for artists, an eye-opener, or a barrier-breaker, that brings us new land to explore.” — Matias Nordin
During the day, Matias Nordin is a senior data scientist. At night, he becomes a digital artist that explores the intertwining of human and machine. He is based in Floda, Sweden, and his primary interest is to discover new meaningful ways of collaboration between the (human) artist and Artificial Intelligence (AI). Truly, his work goes beyond this. In fact, we could say that his artistic and research interests lie between Computational Creativity and Creative AI.
Matias’s work shifts from exploring the outputs generated by algorithms that mimic realism and different artistic genre, to finding ones that are surreal and ethereal. His effort of making new meanings and content to emerge is supported also by the work of “controlling, or communicating, with the AI in an attempt to simplify collaborations between human and machine”.
A chat with him brings back old questions. What is creativity? Can a machine enhance human creativity? Do we need technology to be more creative? It also pushes forward the concept of an “autonomous” machine capable to be an artist on its own right.
Beth Jochim: Can you tell us a little about your background and how did you start to get into the field of AI Art?
Matias Nordin: I have been interested in computer graphics since early childhood. I grew up with computers, but without software, and programming became a natural way of entertaining myself. If I wanted something done, a game to play, music to be played or graphics to be created, I had to make it myself. Add to this some sort of laziness, and the idea that the computer should be able to do the boring parts for me. With time that notion also became my whole career, as a physicist/researcher/problem solver/machine learning expert: to avoid doing boring stuff by writing code that does it for me. I think many within machine learning recognizes themselves using that motto. Creativity remained an important part of me throughout my career, and I’ve always loved to explore and test different ideas, mainly for my own amusement. With time I found myself spending more and more time in front of the computer, also in the evenings and I decided to pick up painting (mainly watercolor, also some acrylics) in an effort to get away from the computer. During this time I was working as a researcher at Stanford University and kept using machine learning and image analysis tools in my daily work. I had an idea that I wanted to learn more about themes and composition in art, to develop further as an artist and for me a natural approach was to start classifying paintings using AI. Soon I realized that my AI tools could be used to paint as well. It was a very interesting time in Silicon Valley and I probably would not have gone in this direction without the inspiration I got from there, it was a vibrant mix of on the one hand, artists, painters, musicians and on the other hand talented programmers, latest tech and a general exciting startup feel. I left the US in 2016 to move back to Sweden and kept using my AI art tools as an evening hobby/project until 2018, when I finally decided to resign from the university and focus fully on AI and art. I could simply not let the exciting clash I experienced in Silicon Valley go, it just kept growing on me. Today I work part time as an AI consultant and part time as an artist. Quite ironically, however, I still spend a lot of evenings in front of the computer.
Beth J.: You create “minimal neural networks” that are artistically and expressively interesting. What techniques do you use to generate your artwork?
M.N.: I mainly work in Python, but often keep relying on the set of algorithms I’ve developed during my time as a researcher, anything from old FORTRAN scripts to C to MATLAB to Python. I write my own stuff for processing images. I don’t really know how to use Photoshop, but sometimes I open GIMP to test different filters I’d like to implement. As for deep learning framework,s I’ve used most of them. I really enjoy Keras at the moment for the simplicity, but the choice of framework really depends on what I want to do and what I have around to make it done. I experiment with hardware too, a mix of Raspberries, Arduinos and old buttons from radios, etc. I used to do make music on sequencers and wish to have similar tools to interact with my algorithms, to get into a creative workflow. I find the programmatic approach a bit slow when I quickly want to shape, change, test, tweak and so on.
Beth J.: What is your relationship in the AI Art community? Who is your biggest inspiration?
M.N.: Honestly I can’t really point on a particular person within the AI community that inspires me the most, the cool stuff I see is often find being done by joint projects. In a sense I perceive the AI field at the moment to be a bit of a grab-and-run, but mainly because there are so many opportunities and everybody wants to do exciting stuff. On the other hand there are these larger very interesting efforts that requires more time and resources. But in such projects the ones that really do the work typically won’t get the credit. Its often the company or organization that claims the fame. I do have many people who inspire me, but they are from outside the AI field; researchers, “classical” (as opposed to AI) artists.
Well, now when I come to think of it, there is this one pioneer that has really inspired me and that can be put, I guess, in the AI space: he is a Finnish researcher called Erkki Kurenniemi. He worked at the University of Helsinki in the early Seventies developing electronic instruments and pushed the boundaries of electronic music as well as the instruments themselves. A very successful combination of an artist, a visionary, a talented programmer and an engineer. He was interested in building machines that created music on their own and had an early vision of how AI would change our society. He even went so far as discussing how we, humans, in the future would become “immortal” as all our thoughts, memories and soul would be uploaded to a “cloud” service. I warmly recommend a documentary called “The Future Is Not What It Used To Be” that portraits Kurenniemi’s life, vision and contributions.
Beth J.: How do you envision the future of AI in visual art?
M.N.: A prolongation of an artists expression. AI is today already a creative tool for artists, an eye-opener, or a barrier-breaker, that brings us new land to explore. I do hope it also may become more of a standard tool for artists as a new type of brush or a camera lens and I don’t mean like a style-transfer tool, but rather a companion or a collaborator for artists that suggests new directions to explore. Personally I will keep exploring the possibilities of using hardware to give means of communication for this type of collaboration. I also believe that AI will be an excellent teacher for emerging artists. After all, it is a very effective means of clustering and classifying the works of the masters. The small remaining gap of how to utilize this condensed knowledge, to suggest for example students how to improve in their work, requires a very small bridge to be built.
Beth J.: In your artistic practice, on one side AI is a creative tool and a collaborator. On the other side, AI is an “autonomous” creator that you observe from a distance. In order to be able “ to stay in a fluent creative workflow”, you are building hardware to facilitate the communication with an AI. We could say that your research interest, and work, are crossing the boundaries between Creative AI and Computational Creativity. So, who is the artist at the end of the day?
M.N.: It depends on my projects. In some of them I am more passive when it comes to the actual drawing and release the algorithm to do as they want themselves. This typically however requires many many efforts and tweaks before I like it, so I guess in a way I still am in control. In other projects I take the active lead and guide the algorithms, both by a workflow that goes like “rewriting, iterating, coffee, stopping, testing, collecting results, conclusions, rewriting, coffee, scratching, stopping, testing”, which I do enjoy as a researcher, but also a workflow that goes: “idea-tweak-draw-idea-add-tweak-subtract” with the use of IO between myself and my algorithm, which is more of a direct creative workflow that sometimes is wonderful and sometimes is frustratingly far away when I realize that I need more software to get it done. I hate to break the creative loop and write more code, especially if I am already able to see it through. Oh man, I wish I had more time (and patience)!
“…The lack of consciousness is not a fundamental reason to deny the potential for creativity or even the potential for intelligence.” — Ramón López de Mántaras
Nordin’s background is in physics and materials science, with a work experience as a researcher in numerical methods applied to different areas (i.e., at the Ecole Polytechnique in Paris, in Heidelberg at the German Cancer Research Center, in analysis and evaluation of satellite data at Stanford in collaboration with NASA). His personal interests, though, go from philosophy to logic, from theater to literature and art. This curiosity for different subjects, together with a strong technical background, has allowed him later to develop a very personal vision and artistic sensitivity.
Nordin is not interested in creating programs that mimic brushstrokes or photo-realism, but aims to develop his own style together with his neural networks. He does this by constantly questioning composition, colors and ways of expression and iterating and changing his programs. In this way, he also challenges the concept of creativity and pushes the boundaries between technology and art. He says: “I have worked with many visual problems where I used algorithms to process images. Over the years I have had a rather large toolbox so it was only a matter of time before I would end up there. It is a bit ironic that I started painting to get away from a daily life where I sat and worked with the computer and then started using it in art.” 
The time spent in the United States has made him decide to leave, at least partially, the research field to pursue a career in art using AI and machine learning. As he explains: “Today I work part-time as a consultant in AI and part-time with art. It is a fantastic combination where I keep up to date with new technologies and have room for creativity. That’s exactly where I want to be, where art takes a new direction.” Then he goes on and says: “Often my neural networks don’t do what I want. In the absence of rules, one opens up for possible creativity, but also for chaos and it is not always easy to predict what a particular network will create... sometimes it takes weeks before I see what I have in mind”.
In relation to the work of the neural networks, the feeling of suspension between being a creator and a spectator is probably one of the reasons that pushes Matias to want to build a communication channel between man and machine. The communication with his neural network is a very important part of his work, as Nordin is developing his style with them. This is why he has begun to build hardware, in order make an interface that can facilitate the sharing and exchanging of information, with the hope “to stay in a fluent creative workflow”.
His art is then print on paper and framed, choosing paper as the right medium for this artistic collaboration between watercolors and neural networks.
Resources and References
 From a previous interview appeared on the Swedish newspaper Lerums Tidning
About the author: Beth Jochim is the Creative AI Lead at Libre AI, and Director and Co-Founder at Cueva Gallery. She works at the intersection of technology and arts. She is actively involved in different activities that aim to democratize the field of Artificial Intelligence and Machine Learning, bringing the benefits of AI/ML to a larger audience. Connect whit Beth in LinkedIn or Twitter.
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.