Images generated by
Images generated by CAN

The replacement of human artists by machine generated algorithms

In a rapidly changing digitized world many careers have already seen a death in human occupation, such as quality control checkers in factories that have been replaced by machine generated algorithms. This bring about the question of what this digitization and the soon to be, if not already, redundant nature of certain career paths means for the arts. The art space has seen a sweep of digitization, and today there are algorithms creating art. Does this art reach a perfection and appeal that human made art cannot? And what does this mean for human artists? Will they be replaced by machines? What does this mean for art history? Are these new “innovations” changing the face of our history?

What has been coined the biggest artistic achievement of 2017 took place in an Art and Artificial Intelligence Lab (AAIL) on the main campus of Rutgers University, New Brunswick in New Jersey, USA. Professor Ahmed Elgammal put a new art writing algorithm through a computer on the 14th of February and witnessed how it created a number of images which left him awestruck. In response to his sighting the professor conducted a unique Turing test (developed by Alan Turing in1950 and tests a machine’s ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human) in order to establish how his AI’s digital art ranked against museum-graded canvases (Chun 2017).

His test took the form of a randomized-controlled double-blind study in which participants were not able to differentiate between the computer’s art and that of human artists. His findings indicated that the computer-made images were in fact preferred by the participants, who referred to them as “aesthetically pleasing”. The paper based on his Turing test ignited an unnerving rumour within the art world – the AI was able to paint like Picasso (Chun 2017).

Following the Turing test Elgammal wanted to move the AI away from fabricating existing artworks and instead create new artworks entirely. The training involved feeding the algorithm over 80,000 digital images of Western paintings from the 15th to the 20th century. Utilizing his own image-generating system – Creative Adversarial Networks (CANs) the machine developed an aesthetic sense and learnt how to paint. The algorithm had one main purpose for Elgammal; to create art that had the appearance of being created by a human brush (Chun 2017).

Finally, a time came where the machine’s art met the bar of museum worthy imagery. The machine was creating abstract work however and strayed away from portraits and still lifes. Elgammal emphases the revolutionary nature of CAN in an interview with Artsy “If you feed the machine art history from the Renaissance to the present and ask it to generate something that fits into a style, the natural progression would be something along abstract aesthetic lines. Since the algorithm works by trying to deviate from style norms, it seems that it found the answer in more and more abstraction. That’s quite interesting, because that tells us that the algorithm successfully catches the progression in art history and chose to generate more abstract works as the solution. Abstraction is the natural progress in art history.” (Chun 2017)

What Elgammal is saying then is that the algorithm did that which many human artists would do in the same situation and produced the same sort of striking images that would catch the eye of curators and/or art critics. Creating art that fell under the classification of ‘Dutch Masters’ would simply not be exciting enough to a viewer and lead to habituation or fail to arouse a response to repeats of this stimulus. Plainly put, to make an artwork appealing to its consumers the visual stimuli of an artwork must have arousal potential in order to create a trigger – referred to by psychologists as a “hedonic response” (Chun 2017).

In his text, ‘The Clockwork Muse: The Predictability of Artistic Change’ (1990), psychologist Colin Martindale implies that the successful artist integrates novelty in their art. His hypothesis is that the increase in arousal potentiality neutralises the observer’s certain habituation retort. This escalation in creative novelty, nevertheless, needs to be decreased to hinder negative viewer response. Furthermore, he believed that “style breaks” are embraced by artists as a means of securing unpredictability and more attractiveness for their audience over an extended period of time. Could this concept be used to base an art algorithm on? Elgammal states “Among theories that try to explain progress in art, we find Martindale’s theory to be computationally feasible.” (Chun 2017)

In 1966 Experiments in Art and Technology (E.A.T) was founded by Bell’s engineers, Fred Waldhauer and Billy Klüver including artists Robert Whitman and Robert Rauschenberg. The pioneering Bell Labs project set the foundation that all computer-created art is based on today. The introduction of new computer technology meant that new machine-generated art followed soon after. This included: dot matrix printer art (1970s), video game art (2000s), 3D-printed art (2010s). The exceptionality of Elgammal’s computer images can be found in the happening of these images marking the first time in history that an A.I. has created work without any human assistance or interference (Chun 2017).

The Greek artist Panos Tsagaris’ ‘Untitled’ 2015 work – a mixed media canvas tinged with gold leaf – was displayed at Art Basel 2016 in Hong Kong and utilized as a sample image for Elgammal’s AAIL tests. The artist expresses that A.I. art is fascinating and regards the algorithm as a peer and not a disturbing threat. (Chun 2017)

“I’m curious to see how this project will progress as the technology develops further. How human-made paintings generated by a machine look is one thing; bringing the A.I. artist to the level where it can create a concept, a series of emotions upon which it will base the painting that it will create is a whole other level.” I want to see art that was generated in the mind and heart of the A.I. artist.” he expresses (Chun 2017).

Art critic and historian James Elkins is more sceptical, “This is annoying because [algorithms] are made by people who think that styles are what matter in art as opposed to social contexts, meaning, and expressive purpose. “One consequence of that narrow sense of what’s interesting is that it implies that a painting’s style is sufficient to make it a masterpiece.” He argues that “If human artists were to stop making art, so would the computers.” (Chun 2017)

Images generated by CAN

Michael Connor, the artistic director of Rhizome (a platform for digital art), agrees with Elkins. “This kind of algorithm art is like a counterfeit. It’s a weird copy of the human culture that the machine is learning about.” He does nonetheless feel that this is not necessarily bad: “Like the Roman statues, which are copies of the original Greek figures, even copies can develop an intrinsic value over time.” (Chun 2017)

Elgammal’s algorithm conforms to the same development process as the human artist – “In the beginning of their careers artists like Picasso and Cézanne imitated or followed the style of painters they were exposed to, either consciously or unconsciously. Then, at some point, they broke out of this phase of imitations and explored new things and new ideas. They went from traditional portraits to Cubism and Fauvism. This is exactly what we tried to implement into the machine-learning algorithm.” (Chun 2017)

Having had its first solo-machine show in October of last year, ‘Unhuman: Art in the Age of A.I.’ in Los Angles the algorithm is steadily rising traction as an emerging artist. Elgammal’s algorithm has ample room to grow its career as the coders at the Rutgers lab are able to increase the “arousal potential” of its artworks (Chun 2017).

What is at stake or what human artists should fear is this; the algorithms’ founder predicts that its art will only get better over time. “By digging deep into art history, we will be able to write code that pushes the algorithm to explore new elements of art. We will refine the formulations and emphasize the most important arousal-raising properties for aesthetics: novelty, surprisingness, complexity, and puzzlingness.” (Chun 2017)

Elgammal is of the opinion that this technology will lead to an infrastructure development that supports his “arousing” art. This will include galleries, online auctions, agents and authenticators fronted by another AAIL algorithm (Chun 2017).

Before getting rid of our artworks crafted by humans it is important to consider the following historical occurrence. In 1964 an engineer and computer pioneer at Bell Labs, Michael Noll conducted his own Turing test. His test was conducted by programming a General Dynamics microfilm plotter and an IMB computer to spit out an algorithmic creation of the Piet Mondrian’s ‘Composition with Lines’ (1917). His digital image was projected onto a cathode ray tube and recorded with a 35mm film camera. A duplicate of the print, ‘Computer Composition with Lines’ was shown to 100 people curated next to an image of the 1917 artwork by Mondrian (Chun 2017).

What was observed from all this was a striking trajectory. 59% of the subjects in the test took preference to the computer-generated image as opposed to the Mondrian original. What was even more remarkable is what took place a year later when Noll’s digital art was exhibited at Howard Wise Gallery in New York, (the first-time computer created art was shown in an American art gallery) – public interaction was poor and not a single art work was sold (Chun 2017).

Nevertheless, Noll retained his optimism for the future of digital art production. He wrote in 1967: “The computer may be potentially as valuable a tool to the arts as it has already proven itself to be in the sciences.” And so,the mad scientist invents a machine that becomes more intelligent, more artistic – not a tool to aid artists but to replace them. What do we make of a machine that has the capability to become more human than the humans? (Chun 2017)

But what if I told you that since and even before the inception of digital curatorial spaces such as Instagram, humans have been impersonating algorithm art? From art schools to gallery circuits a prevailing style of abstract painting is coming afore. This style has been called “Zombie Formalism” (basically Neo-Modernism or Crapstraction) by critics. Its characteristics have been singled out as spinoff-ish and pretentious; making use of the vertical format for convenient Instagram hosting. These works that have been digitized and “filtered” in some cases is where the masquerade of human based art lies (Chun 2017). It does however go further than that, think of digital artists who work with gifs, text messages and screen recordings to produce work – all a game of mimicking technology.

This irony then acts as a catalyst for conversations about a near dystopian future and the possible end of culture. Will humans have to paint more like robots? (Chun 2017) The crapstractionists deviating from the norm are already paving the path towards work that is less human and are altering art history to an extent that they might not fathom at present.

Even deeper than all this is the fact that we have come to a moment in time where images have undergone a massive transformation. The vaster majority of images are made by machines for machines and humans are not involved in the process. As Trevor Paglen states, “We’ve traditionally thought about images and the role they play in society as being centered on a human observer—a human looking at an image. But we’re now at a point where most of the images in the world are invisible. What I’m talking about when I say, “invisible images” is the advent of autonomous seeing machines that don’t necessarily involve humans, and that’s the majority of the types of seeing that’s going on in the world now.” (Abrams 2017)

Can a machine fulfil the functions of being an artist when being an artist is so much more than creating art? Being an artist intrinsically means being a teacher, being an activist and commenting on and critiquing the times that they are situated in. There is also the philosophical and art historical question – what is an image/artwork if humans are not needed to create it? Not to mention the political and ethical questions linked to means of observing as they are also a means to forms of power (Abrams 2017). These factors must not be dismissed so easily and the divide is real and apparent whether humans choose to acknowledge it or not. Humans are slowly being replaced by machines but a machine could never be an artist and fill the scope connected to this classification. It fails to inhabit a human form; for now.

Share This Article


Suggested Posts