Machine learning is challenging the concept of copying. Whether algorithms produce creative outcomes or solely reproduce and therefore copy data remains to be scrutinized and questions our definition of copyright.
The copy is usually perceived as a downgrade from the original, moulting value and identity. Is copying merely a negative force or are there potential realms where copying can positively fill arising gaps and expand the original?
Mother Nature, likely the most origin force on earth, has suffered significantly from human impact in the last centuries. Humans’ interference has irretrievably damaged the diversity of animals and plants and has ultimately altered evolutionary processes. While humans and predominantly their machines have done harm, is there a chance machines may also cure the aftermaths? Copying might constitute an answer to supporting nature’s reproductive forces.
Can we create diversity by copying? In an infinite cycle of copying, who keeps track of the original? And will copying ever be perceived as an expansion rather than supersession?
This curated body of work has been generated by a Generative Adversarial Network (GAN), an unsupervised machine learning framework. The GAN’s discriminator and generator neural network compete with each other in order to generate output that reflects patterns or more broadly the generative variance within the input data.
The input data comprises a set of Bonsai trees, a cultivated modification of nature that merely mimics the shape and scale of original trees. Through interminable mutations and human interference, a new species has been created from which the original is hardly discernible. To fight the oblivion of the original, the yield outcomes of the GAN framework are overlayed by the original input data. There virtually exist two layers, a visible of the artificially created outcome and an invisible of the original, which is however only visible through an Augmented Reality application.
Besides the generated trees, the typeface of the project has also been created by the GAN framework trained on a big set of existing typefaces.
During training, the learning progress is measured by the loss function. The evolution of the loss value represents the remaining distance on the learning path to the training objective. This means, the smaller the loss value (LV), the more developed the model’s abilities. One usually expects the loss value to drop logarithmically over the course of time measured in minutes (MIN) and the number of epochs (E). An epoch defines a single training cycle through the input data and training involves several runs.
Artificial algorithms are becoming an indispensable tool in everyday life. Treated as ubiquitous and trusted technology, Machine Learning is assuming shape in more and more tangible forms: Consumer-orientated primary points of interaction such as Siri, Google Home and Alexa, took over an imperative role in our pockets and homes. Besides these user-friendly applications, Artificial Intelligence is pervasive in financial trading, online bots and dozens of background applications specialised in highly complex tasks.
In the context of creativity, the interplay of human and machine elicits the question of authorship, copyright and originality of the generated output. One may claim that copying is the origin of everything. Without the ability to copy, children wouldn’t be able to learn how to walk or speak. Nevertheless, the copy in Western context is associated with slightly negative impressions and evokes questions around originality and authenticity.
The discourse on Artificial Intelligence and creativity is not a newly arising subject. As early as the 60s, computational creation experienced an unprecedented impetus. This new perspective on machines as artistic creators evoked questions about the relationship between human and machine. Since the emergence of this debate, technology has improved and developed significantly, wherefore the amount of sovereignty of the machine is evaluated increasingly.
This development is generating an area of tension in which mankind finds itself confronted with the first invention empowered to create unpredictable “creative” outcome.