The article, Amplifying The Uncanny, by Broad, Leymarie and Grierson offered an insightful knowledge of how machine learning algorithms can be optimised and used to produce uncanny results.

Machine Learning & Generative Adversarial Network

Machine learning is a system where pattern data are used to predict future data or other outcomes of interest. It is understood as an automated process of optimisation, where an algorithm processes data and finds a set of parameters that best ‘solve’ a pre-defined objective function. DeepFake, a software that produces images that can trick the human eye into believing that they are real; it utilises machine learning, in particular, the Generative Adversarial Networks (GAN) framework to produce a sample of realistic images.

Next Great Movement in Art or Zombie Art?

Generative Adversarial Networks have then been widely used in the production of art and is seen as the next great movement in art. However, it has also received several critiques. Firstly, the production of new artworks by the dataset of existing artworks is seen as dull and obvious as Hassine and Ziv called it the “Zombie art”. Another criticism is that generative artwork relies too heavily on deep neural networks to produce mesmerising samples, without any meaningful framing of the works being presented by the artist.

Argument

Perhaps, the reason why Hassine and Ziv label generative art as “zombie art” is the problem of how it is very difficult to value generativity. One can’t see at a glance whether it took time or skill to create, similar to photography, one may think is just a press of a button to produce art as compared to paintings or sculptures which are seen to require more work. Nonetheless, from the experience in making my first sketch, generative art is definitely not just a press of a button and it does require creativity in the process.

Next, even though machines are heavily relied upon to automate some aspect of the generativity, it does not mean that it is not significant or meaningful as compared to other fields of art. Rather, I argue that it is actually a new form of creative outlet that artist could utilise in producing art pieces.

“Being Foiled” & the Uncanny experience.

Expanding on GAN, one of the explored artworks in this paper was “Being foiled”, which reverses the outcome of DeepFake algorithms. Instead of producing images that look imperceptibly realistic, “Being foiled” invert this objective, by optimising the system to generate what it sees as being fake and turning the image into complete abstraction. As a result, it amplifies the uncanny nature of these machine hallucinations.

The ‘Uncanny‘ is referred to as a psychological or aesthetic experience that can be characterised as observing something familiar that is encountered in an unsettling way. The uncanny effect is usually seen in objects that have a likeness of human qualities, whereby it is close to human-like but not human; as a result, it becomes creepy. For instance, a robot that looks very similar to a human, which invokes familiarity yet an uncomfortable feeling.

The author illustrated his point of uncanniness by picking an iteration of the model whereby the uncanniness is potentially most amplified. Indeed, just by viewing the sample, it does provoke a feeling of uncanniness. It is familiar, yet it gives a sense of unease. Relating this to generative art criticism as dull or monotonous repetitions, “Being foiled” actually shows us how generativity can invoke human perceptions and feelings. In this context, it provokes a sense of uncanniness to the audience. 

Hence, the question is, how can we incorporate intangible emotions into our sketch that can be understood visually and consequently make generative art less banal?

Closed Systems: Generative Art and Software Abstraction

By Marius Watz, 2010

With the rise of computational technology, art is going through an evolving process with the combination of the principles of unpredictability and the purity of logic as Marius Watz describes it.

In his paper, Watz touches on the notion of exploiting the complexity produced by a computation that could not be made by human hands. He elucidates that generativity is a useful strategy that allows the author to harness the power of a computer to generate an infinite series of possible outcomes. However, to perform such a process, the author requires technical skills to be able to understand computer code and perform reverse engineering to achieve a desirable outcome. Therefore, as Watz contends, generative art requires both technical skills and aesthetic intuition for the artist to be able to express their art through the manipulation of computation. This resonates with me, as someone who sees through the lens of both art and programming, as both are vital ingredients to produce an interesting computational generative artwork. Moreover, to a certain extent, computational skill is an art because it requires critical thinking and ingenuity to create objects of beauty.

On the other hand, computational aesthetics also has its constraint. As pointed out by Watz, simulation is spontaneous and organic, which can only be replicated by computers with the explicit encoding of how the data is processed and even the most experienced programmer will encounter unexpected results or errors. Nonetheless, programming error can lead to discoveries and new learnings. Indeed, as a programmer that is still learning, more often than not, I gained valuable information from errors. For that reason, I see mistakes and errors as a beneficial process of discovery, learning and improvement.

Besides that, Watz also made a comparison with interactive art and generative art. He elaborates by drawing a distinct difference between generative art and interactive art. Generative art operates based on an autonomous system with infinite outcomes based on the algorithms, while interactive art utilises a feedback loop of interaction between a system and its users. Based on this definition, I do see the distinct differences and mutual exclusiveness in their respective art forms. The way I see it, both arts require controllability through a set of rules. In generative art, the algorithmic rules create art, and for interactive art, the rules of human interaction create art.

Beyond that, Watz has also postulated that generative art has reached a level apart from computation means. While Watz has embraced the complexity of computer systems, he does agree that generative art is not about the computer itself and would be a mistake to think that generativity is only capable of being expressed in pixels. Thus, generative art can come in many possible outputs, and I certainly do agree and believe we should expand and explore our creativity of generativity in many various possible outputs, and not just limit our observations of generativity to solely one medium.