Original article was published on Artificial Intelligence on Medium
How do you stand out?
A question every artist has asked themselves at some point in their lives, standing out can be quite a tricky business, especially given the influx of styles already developed and spread across the World Wide Web. Art is a competitive market, and experience has shown me that becoming a full-time artist is easier said than done. How then, does today’s featured artist tackle making a name for herself in the ever-expanding genre that is AI Art?
Russian AI artist Helena Sarin graduated from the Moscow Civil Engineering University, and through her love of the arts, as well as her being a software engineer working with CycleGANs, she opened up to the possibility of training one using data sets she created herself. In doing so Sarin employed a way of protecting her generated artworks against the increasingly similar art pieces being generated by those with access to better technologies (like BigGANs) — by ignoring the race to better computational power and photorealism completely and working with her own original data sets. Bigger isn’t always better, and artificial intelligence art was for some locked behind a hefty pricetag — as the AAA software does not come cheap. Sarin has shown that just like in traditional art, it is not the tools alone that create great artists, but creativity and the will to express it.
The elements used in the creation of this series of works are visible — printed text and tree silhouettes — instilling quite an interesting stamp/transfer effect. Since their creation in 2018, Sarin has continued training the model and documenting the results, compiling them into a collage of works that look like traditional woodblock prints.
Will AI art be a never-ending computational arms race that favors those with the most resources and computing power? Or is there room for modern-day Emil Noldses and Erik Heckels who found innovation and creativity in the humble woodblock, long after “superior” printmaking technologies had come along?
Through trials and tribulations, the collage serves as a pioneering achievement to AI artists looking to develop a new style with GANs on a budget, or traditional artists that are interested to try a whole new medium entirely. Sarin has opened up to stylisation in neural art — much like traditional art after the creation of the camera — a commendable practice in a timeline where photorealism’s influence continues to spread across different media (live-action films, GANs, video games, etc.)
But how does AI learn to make art?
It trains, just like us! In talking about Sarin’s ‘woodblock’ works, let us pull back a little bit to see how Generative Adversarial Networks learn to create art, so that we may understand and appreciate the process each of her (and other AI artists’) artworks go through.