By Morven Watt
If Michelangelo suddenly appeared in 2020, he’d be amazed by the number of people crammed into the Sistine Chapel to marvel at his works on any given day. He’d probably be equally amazed at the volume of that art that is produced around the world, every minute of every day.
Apps like Instagram, Patreon, and Etsy have made art sharing not only possible but profitable, too. You only need to check out Netflix or any video game to see the endless possibilities afforded to us by digital art.
Now, new facets of data science and machine learning are creating more artistic possibilities than ever before. With tech that can create art, we’re about to enter a whole new era of artistry.
Let’s start by looking at the science. CycleGANs are a type of generative adversarial network that doesn’t need paired images. GANs are algorithmic architectures that use two neural networks pitted against each other to generate new synthetic instances of data that can pass as real data.
In regular image-to-image translation, you would have paired data sets. In a CycleGAN, you have unpaired collections of images from different domains.
In regular image-to-image translation, we see things being done like changing landscapes from summer to winter (or vice versa) or a painting to a photograph. The problem with this is limitation—the data sets can be challenging and expensive to prepare, or simply don’t exist.
What’s fascinating about CycleGANs is the sheer volume of possibilities that it affords. Aside from being fun to play with (like you can do here, here, and here), it also offers a glimpse into future possibilities with this type of tech.
Movies and video games are some of the first uses that come to mind for obvious reasons, but there’s a galaxy of possibilities out there. From authors who want to add creative imagery to their books, to new ways to immerse kids in art at any age, to the ability to create your own art if you’re not artistically inclined, CycleGANs can offer something for everyone.
Impressionism, pointillism, surrealism—and now AI-ism? Since humans do factor into the creation of this artwork, it still has the element of creative control from outside input, but the creativity of the machine cannot be downplayed. We’re essentially seeing a new genre of art being created.
Play around with Artbreeder and create a mini video of changing landscape with a castle that crumbles over generations. Play with Nvidia a create a sea of burning trees springing from the ground. While you have some input, there’s still a randomness that you can’t predict with these programs. Even watercolor, considered to be a “loose” and somewhat unpredictable medium, still has more control from the artist.
For the naysayers out there, digital art has the same issue. Plenty of traditional artists think the only “real art” can be done with a canvas and paint. But you only need to look at some of the pieces on Artbreeder to see the level or artistry this tech is capable of. And for every naysayer, there will be a host of artists, writers and creators thrilled that they can see characters and concepts brought to life without the need for middlemen.
Since this tech gives control back to creators, perhaps we can look forward to a future with fewer remakes and sequels, and one with more innovative content.