The Problems With Open Source Code

about 1 month ago by Sophie Stones

The Problems With Open Source Code


Last month, an AI generated artwork created by 3 French students under the name Obvious, sold for $432,500 at a Christies auction, marking a milestone for artificially generated artwork. The portrait, created using GAN (Generative Adversarial Networks) and named Edmond de Belamy, as a nod to GAN inventor Ian Goodfellow, has raised important questions regarding the origins and usage of codes.

In layman’s terms, GAN used in such a way is a kind of complicated paintbrush with mathematical parameters. Learning from a large amount of input data, it is able to use that knowledge to produce new results, with the end product being a result of an arduous process of carefully selecting input data, tweaking parameters and then sifting through the results to find the best example. To put it simply, stating that the AI created the painting is a rudimental way of viewing Edmond de Belamy, rather both people and technology work together to produce something that is an amalgamation of past work. A fact that was negated both by tabloid papers and in order to market the painting as something spectacular.

What further complicates this case is that the code used to generate the painting was mostly the work of another artist and programmer, 19-year-old Robbie Barrat. Barrat, who released the open source code, is an artist in his own right, renowned for his AI generated landscape and nude paintings. Barrat himself has stated that he holds no grudges towards Obvious, however others in the digital art space aren’t so generous. Jason Bailey, Digital Art Blogger and owner of Artnome has stated “it’s almost weekly in digital art that someone takes open code and tweaks it and sells it… there’s a lot of stuff you can do that’s legal but that makes you sort of a jerk.”

So, are Obvious jerks or are they just misunderstood?

A large amount of the backlash towards Obvious has been down to their relatively new presence within the Digital Art community, with many thinking they have coopted what they feel is a community, for monetary gains. In an interview with Hugo Caselles-Dupre, the tech lead from Obvious, Bailey states that “for artists that have worked for five to ten years on AI, it’s frustrating to see someone come to GANs with no artistic background, use a stock GAN and end up at Christie’s.” This frustration isn’t just down to the money made but what perception this newly famous painting is giving. The members of Obvious aren’t experts with the GAN paintbrush and this shows, leading Barrat and other AI artists to worry about the impression outsiders are getting. It is confounded by a belief that the marketing surrounding Edmond de Belamy, including the tagline “Creativity is not only for humans”, is vastly overrating AI’s role in creating the work and is misleading those not familiar with GAN technology.

The question of who deserves credit and whether people should use open source code for profit, has been raised many times, with users on Quora and Stack Exchange asking whether it is cheating, however no real conclusion has been shared. And so, whilst Obvious admit Barrat deserves credit for his role in creating the first commercially sold AI painting, he won’t be receiving any of the profits just yet.