AI and the Intimate Art of Hannu Töyrylä
Explorations of AI Art — Episode 21
[This interview has been previously published on Cueva Gallerys’ blog on July 6, 2020]
“ I always had a need for the kind of self-expression that is associated with art.” Hannu Töyrylä.
Hannu Töyrylä is a visual artist based in Helsinki, Finland, who makes video, interactive installations and art prints using Artificial Intelligence. His work revolves around recurring themes, such as the contemplation of his emotional, psychological, and spiritual world, his life experience, and the complexity of our times. As he explained to me, his interest is not in machines as autonomous creators, but in approaching AI as an augmentative tool of expression.
Töyrylä speaks of himself as a non-AI purist. Indeed, a distinctive feature of his art is the mix of manual work and image processing code, starting for example from photography. His unique background combines a M.Sc. degree in applied electronics and telecommunications at Helsinki University of Technology, a PhD in Jewish Studies at the Theological Faculty of Åbo Akademi University, and an artistic path that spans from music to literature to visual arts.
His art moves from a visceral, almost existential need for self-expression, and it is precisely through art that he seeks to connect the constituent components of its identity with the world. Building his tools, and the materials from which to draw inspiration (i.e., small datasets), helps the artist to connect with his work, accentuating his feeling of belonging to it.
Moving from a human-centric conception of art, Töyrylä develops a very personal approach to neural art, alternating a transformative process with a generative one and finding a balance between discovery and control. Trying to develop a gut feeling with his neural networks, his practice leans toward exploration and experimentation.
For instance, in City views from a dream, the artist creates human environments with a note of desolation, bringing to life a dreamlike image of the city and mixing elements from the surrounding nature. In Keeping the chaos out, photographic images maintain their center untouched, while, all around, their edges and parts are distorted and twisted. Lost years investigates the loss of the artist’s position as an expert at Nokia, and the impact of this experience on his identity. Recent, instead, is an experiment where the artist’s intention is to reduce images to their bare essential. The realization of this attempt implies to interfere with the neural network in order to reach a blurred effect before a clear image can emerge.
Töyrylä’s art is intimate. The artist’s inner world is enriched with the scholar’s thought, and the curiosity and technical skill of the engineer. His uniqueness as a man becomes a symbol of our humanity made up of memories, feelings, experiences, thoughts and actions. The artistic practice happens to be a moment of reflection and research, in which the individual meets plurality in a mutual exchange, and where Artificial Intelligence proves to be an augmentative tool for self-expression.
With us, the artist has discussed his work and vision, and shared two new pieces that will be part of his upcoming exhibition at Pesula Galleria in Sipoo, Finland, in September 2020.
Beth: You are ‘an engineer turned humanities scholar turned visual artist.’ How have the technical and humanistic paths helped you shape the artist you are today?
H.T.: My background is of course essential for my art. Turning seriously into AI art has required full use of my technical skills and allowed me to explore various ideas on my own. Humanities, then, has supplied ideas and depth into my artistic work.
These paths are linked to diverse aspects of myself, all present from the beginning. Of course, in real life these aspects were actualised in a sequence. When working with technology, the other aspects were mostly latent. Working with humanities research was a step in healing oneself, becoming whole. Now, the three aspects are in balance. I can move between a poetic expression for what cannot be expressed in rational terms, and applying technology whenever my expression needs it.
Beth: What is the role of art in your life?
H.T.: I always had a need for the kind of self-expression that is associated with art. It comes from my home. My parents were writers. As I grew up, art, be it literature, music or visual arts, never was for me only an object for passive contemplation or consumption. It usually triggers in me a need to do something myself.
Now if making art is about expression, experiencing art is, perhaps, about perception. For me this creates a circle, experiencing and expression work in tandem. To a large extent my self emerged from this. Art becomes self-development, both in a narrow, technical sense of developing as an artist, but also in a wider, existential sense.
Beth: What does creativity mean to you?
H.T.: I am not really interested in defining creativity, it would only take attention away from the activity itself. Distance myself from my work.
In mythology, the creation of the world is often seen as bringing order out of chaos; real chaos in which there is no form and nothing to perceive at all. Human creative work then, although it happens on a smaller scale, is not so different. An artist has to struggle with a void which hopefully should give birth to a work, and a chaotic universe of materials, options and influences. Making the puzzle pieces gradually fall into place, so that meaning and aesthetic value emerges.
Beth: How do you approach AI in your work and how did you get interested in neural media?
H.T.: Working with AI art today almost by nature requires attention to both the technical and the artistic side. It feels mandatory to both build my own tools and rely on my own visual materials. This dichotomy creates a tension, like when I started writing my own GAN code and had to spend several months with very little artistic work done. Then, at other times working with both elements at the same time can be a great experience: modifying the code and working on the image materials at the same time to get everything match perfectly.
I was interested in AI already in my youth, even took a course on AI at my university in the early 80s, entertaining ideas of creative use already. Thus, the option of using AI was already there, waiting for its time. It came in 2015, when I was getting more into visual arts and looking for a software based technique to simplify and transform photos into something more abstract. It was then I came across with deep learning and its image applications. Neural style transfer had just been released. It could not really do what I wanted but I persevered, went on experimenting, following the field for relevant developments, more and more turning into modifying code and finally writing my own.
Beth: Your art is often a delicate balance between ‘transformation’ and ‘generation’. Can you guide us into your creative process?
H.T.: Deep down, it is a question of how, when working with AI, to make works that look and feel my own. Not just any works that look like art. With transformative techniques, the artist has more direct control on content and style. Generation tends to lead into exploration: you have to look for the right scene and style instead of setting them up yourself.
My early work was mainly transformative: taking photographs and using style transfer to make collages which were clearly mine: often self-biographical with an added dreamlike quality. Likewise when I used pix2pix to make virtual prints out of thread frames, or applied virtual scratches to my photos. With such techniques, the composition of the work remains under my control throughout the process.
It is different with generative, especially the kind usually used with AI and GANs. The role of artist leans towards exploration. To feel more on my own, I decided to write my own GAN code. Borrowing ideas but also trying out my own crazy ideas. Trying to develop a gut feeling how my GAN is behaving. Storing good models to use as seeds for subsequent training with different image materials.
With a GAN, however, you can only achieve so much with the code. The larger part depends on the training images. Selecting the images used in training is the key. Because I use my own photos, my sets are anyway small. I may use a bit larger set for initial training, then reduce the set as the training proceeds. Discard images which tend produce undesirable visual elements. This way, while I cannot zoom in on a specific image, I can steer the process to produce a specific world of images, all images sharing a coherent set of qualities.
I am not an AI purist. I prefer to mix other phases into the process, both manual work and image processing code, rather than trying to build a purely AI based process. Just now I am experimenting in combining images produced by my GAN with film photography, and it looks promising, close to combining best aspects of both.
Beth: What is the artist’s role in relation to the augmentative power of AI? What is the value of AI art?
H.T.: I would say the most essential thing for the human artist is to have an inner drive for artistic expression and an openness for self-development. What then remains to do is to develop an intuitive grasp/feeling of how the AI perceives images and how it can be used.
My conception of art is human-centric: need for self-expression is a basic human need. An automated production of art-like objects might satisfy the needs of a market, but there would still be human artists with an inner drive towards self-expression.
If the artist wants, AI can function as just another tool or media, in a chain of developments from painting to printing, photography, video, computer generated art and now AI assisted art. In this chain the role of artist gets more indirect. The importance of physical involvement and skills diminishes, while the requirement for vision, perception and intuition gets more into focus.
Then, the human artist is the one with a need for expression, guiding the process according to her vision, making choices and determining when the work is ready. Note also, that AI can often be used as a part of a larger workflow: the human artist may, for instance, continue to use other means to work further on whatever the AI has produced.
Beth: Can you tell us about a project that has been of particular importance to you?
H.T.: Usually my projects alternate between being content driven and process driven. I can prepare a new image set in order to create a new world of images, like when I photographed various household objects together with their shadows to create What you see, a series of semi-abstract images. Alternatively, I get a new process related idea, like when I inserted median filters into my GAN code to see how it would affect the created images.
The story of my solo exhibition Faces, memories is different. It started from technical experiments, training a VAE [variational autoencoder] to create human faces. I ended up with a gradually evolving display of somewhat averaged facial features, which also started to adopt the features of a spectator if present. Added some post-processing get a distinct, a bit hazy and remote look. This was an interesting exercise into human facial features, but it lacked a context, which came to me afterwards. A small table with an old telephone and a photo album, just like we used to have at home in the 70s. Instead of the mirror we used have on the wall, I now placed the display of the gradually changing faces. Not unlike our conception of the future in the 70s.
I had arrived at a representation of how we communicated back in then 70s, how we collected our memories, and how we expected this to change in the future. The next step was to tackle the present, the reality how the future turned out to be: most of our originally private communication as well as storing our memories happens in the semi-public sphere of social media. Another display, like a giant tablet, showing constantly changing content, both real and computer generated, with a small papiermache doll looking at it.
So, the exhibition turned out as an installation demonstrating how our life has changed from the 70s to this day. Quite different from the art prints I usually do, although I managed to hang a set of prints on the walls, too.
Beth: How is your relationship with the public and with the local community?
H.T.: For those who are interested in AI art and especially how it is done, I have an FB group I started already in 2015, where I post quite frequently about my experiments and activities. It is definitely not an art show, but rather an ongoing blog about experiments and techniques, regardless of they turn out successful or not, with some actual works thrown in at times.
With regard to the local art community, I am just another grass roots level artist, using techniques that very few know anything about, and most people do not care either. But they have accepted me as a professional artist, I have been able to participate first in joint exhibitions and then keep my own exhibitions too.
Aside of my artistic work, I want to disseminate understanding and know-how about AI art. This has brought me in contact with artists both interested in applying AI in their work and enough technical background to start doing so. Still, both among the art community and the general public, AI art is rather unknown and misunderstood. So there remains work to do.
Beth: How do you see the field of AI art progress over time?
H.T.: I usually do not think much of this. My conception of art is a human activity proceeding from a need for self-expression. As AI develops, it may be difficult to maintain such a balance that is essential in my work. I can see an industry arising from computational creativity, serving human needs for consumption, but it is difficult to see how it would relate to the human need for self-expression. Sometimes I think that art as a human activity may distance itself from technology altogether. Thinking of this right now, I guess what is needed to maintain the balance is enough artists, unafraid of technology, keeping up the creative human use of intelligent machines in the face of a growing industry of automated creativity.
To follow Hannu Töyrylä:
Website: http://liipetti.net Twitter: @htoyryla Instagram: @htoyryla
References and Resources
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.
