About
In the aftermath of French people "stealing" code to generate art and making the equivalent of 20 years of PhD salary, I thought it would be a shame not to try it too.
The animation above is based on 1,500 images generated from a pre-trained model by Robbie Barrat. My only contribution was writing some dodgy code to navigate the latent space of the trained model. For much more impressive animations - recent work (as of November 2018) can be found on Twitter.
How it Works — A Layman's Intro
The model used is a GAN (Generative Adversarial Network). It learns a representation over "all possible" landscape paintings by observing a limited set of training examples.
This process mirrors human generalization. We've seen many dogs, but when we see a new dog, we aren't surprised — we have a representation of "dog" that accounts for unseen variations. Generative thought allows us to imagine unseen examples from limited experience (like drawing a dog we've never seen).
Another example is imagining a pink elephant. We combine ideas (pink + elephant) to create a new one. While current algorithms need thousands of examples to construct complex representations, humans require very few.
Thoughts
Will AI replace artists? I worry that in asking this, the artist becomes the metaphor for the technology, when a better one might be the paintbrush.
If we view art as a creative process rather than just the objects produced, then exploring how to use a tool is art — even when that tool is wide different from what art exploration has been before.
"I never made a painting as a work of art. It's all research." — Pablo Picasso
P.S. After tinkering with the code to generate paths in latent space, it sort of feels mine. (Please DM me for prices)