Category: Research

What is generative art? Or rather, what is art? A question that begets the foundation of discussions for creation of art. “What is art?” varies for each individual; to some it is a product of a free reign of intellectualism. While for others, it might be the first step to a constructed discussion of aesthetics and the foundations. Maybe even both even for others.

But moving deeper than just the broad scope of art, instead of going about the countless theories and form of art, here we explore a certain focus of arts known as generative art. As the article explains, “Art created by means of an apparently autonomous system or process is most frequently referred to as generative art, a realm of digital art practice that has boomed since the start of the 21st century. In fact, the growth of generative digital art has been so robust that, for many people, generative art and computer art have become synonymous terms. In this chapter I hope to show that generative computer art is in fact a subset of the larger field of generative art. It will be seen that generative art can leverage virtually any kind of system, not just computers, and that it in fact is as old as art itself.” 

Now when the term generative art is mentioned, the natural connections to the term would be the root words, generative, and art. Put simply it’s a form of art that is generated rather than created by the manual hand. To do so one requires some sort of system that might accept some sort of input, while creating a unique output. And these sort of input/output systems come in all sorts of varieties, including but not limited to: computer/electronic music, computer graphics and animation, demoscene and glitch art, open source digital art, industrial design and architecture. Across all these forms, there are certain similarities seen such as: generative art is art that uses randomization; generative art is art that uses genetic systems to evolve form; generative art is art that is constantly changing over time; generative art is art created by running code on a computer. These however, are more restricted by the specific form of medium they apply for, rather than generative art as a whole. 

So, what actually defines generative art? To answer this we need to return to the initial topic of “what is art?”. Being a form of art, to answer the definition we need to have a solid understanding of the purposes and function of art, where it comes from and how it is explored. The author of the reading hence gives their definition:

“Generative art refers to any art practice in which the artist uses a system, such as a set of natural language rules, a computer program, a machine, or other procedural invention, that is set into motion with some degree of autonomy, thereby contributing to or resulting in a completed work of art.”
(Galanter 2003)

The idea of generative art would be as explained earlier, through some sort of system a result is created outside the manual control of a human, normally through some sort of automation. However, this initial definition is not I something I would agree with since it’s vague and does not give enough focus onto the idea that it is a work of the system, rather than an assisted work done by the artist. To my delight, the author did create an updated version of the definition several years later:

“Generative art refers to any art practice in which the artist cedes control to a system with functional autonomy that contributes to, or results in, a completed work of art. Systems may include natural language instructions, biological or chemical processes, computer programs, machines, self-organizing materials, mathematical operations, and other procedural inventions. ”
(Galanter 2008)

This definition is clearer on what I agree to be a definition of generative art; control relinquished by the artist to the system, to let the system run on their own creating a wider array of possibilities, limited by what the artist defines beforehand, unlike the previous definition which implied a system that the artist is able to use as a tool restricting their own undesired effects.

Next, the reading explores the idea of complexity in generative art and how it grows. Generative art can be both highly ordered as well as highly disordered, as with the nature of an artist giving control to a system, they can decide how much control to give, as ultimately not all end results are favourable, but highly dependent on the type of art being made. In some sense you can even say if your supposedly chaotic looking generative result out of sheer randomness creates very neat and orderly straight lines, even if you are willing to accept whatever result it would give you, this might still give you doubts. After all, if your generative art doesn’t look generative at all, would you still present it as generative art?

How does this mean anything in my work however? In the process of a game where the final result is not meant to be fixed, there are many techniques I can pick up to define how the final product could be, whether it is depending say a certain level of growth into the work, or applying restrictions. As an artist I am ceding control to the generative art system. I set the rules but I don’t decide the final product, directly. From a gameplay point of view, there are several paths to consider; should I want a very direct non-linear narrative, it is possible to setup enough restrictions in a generative system that no matter how it varies, the final product will be the same. Or I can make it that due to how much it can branch out, having two identical results is extremely unlikely. Simple examples would include Solitaire; each game can have a varied amount of progress but the final product is almost always the same. But that is not always bad, a game of solitaire with a varying finish state would simply not be the same game, and might not give as much as a desirable effect as solitaire as is afterall. To take away from this, as an artist I choose what freedom the system has, and what choice I make should factor from my intentions, leading back to the original question of “what is art?”

The reading begins with explaining that in recent years, machine learning systems grew excellent at producing images, with major focus on creating human faces that seem very real, resulting in possible controversies with technology such as deepfake.

Deepfake is a technology of generating faces by referring to existing images, to create brand new faces which cannot be found through image searches due to them being unique, which leads to the reading further explaining machine learning where the machine finetunes information to develop better or more realistic images. In which leads to the concept of the uncanny valley, which explains how the closer something resembles human, the more welcome it would be, except for particular range of resemblance which seems creepy.

While uncanniness is usually avoided, the reading explores pushing uncanniness to the limit, exaggerating certain facial features to make them look unpleasant, even pushing further to extending to a group of people. The takeaway I find from the reading is that while machines can extend a process, it does not know the limits and can push beyond what we find desirable, for better or for worse.

What does this mean for us? It’s the human’s physical touch and control that adjusts generative art to something that is useful for our purposes. A system that runs without control is likely to create a disaster, so from an artist’s point of view, we need to learn what the limits should be and how to control them to reach a parallel with uncanniness.

Marius Watz describes generative art as “a computational model of creativity combining principles of unpredictability with the purity of logic”. He explains that from results that cannot be created through human hands, the nature of work is complex, even unpredictable. Generative art often takes inspiration from nature, but there is a struggle between natural forms and mechanical forms; what artists balance is how it can appear natural yet refined through their control, creating a chaotic work that flows in a direction they wish.

Put simply the artist allows the computer to create countless results then handpick possibilities that suit their demands, controlling the process that it remains random yet able to only produce results that they find desirable. As long as the results are desirable, the more chaotic the variations of results the more preferable. This means they do not heavily restrict what can possibly happen; they just control the general direction of the work’s flow. Ultimately a level on unpredictability is a boon, the bane is when it goes fully out of control. What they want is an aesthetic mess that conveys their intentions; controlling so is the skill of the artist.

As a reminder, generative art isn’t a movement, it’s just a strategy for creating art. In fact generative art techniques have been dated in human history, from early days of medieval art techniques to later random computer generated screensavers. Artists have always employed chaos as a technique to keep their art lively, and produced unexpected results which are later studied and capitalised on.

Nowdays, generative art is actively used as a concept in creating works by artists. This is interesting for me, for it means that art can be presented at a partially developed stage, and results can be flexible, sometimes even affected by the audience or users of the work, creating more personal results. This has been used often in interactive media, such as games where the input of players may affect how the story or events turn out; the artists basically create what is known as a sandbox which allows many possible unique outcomes, many of which even unplanned by the artist themselves, hence creating a large range of experiences, allowing different people to share how they interacted with it in different lights.

 

Clip taken from YouTube, featuring TPP against the Champion

Twitch Plays Pokémon (2014) 

Beginning on 12 February 2014,  Twitch Plays Pokémon was a social experiment that ran for 16 days with an estimation by Twitch of over 1.16 million participants, with peaking simultaneously at 121,000, while with a total of 55 million views during the experiment.

So what is it?

Twitch Plays Pokémon was a stream of a Twitch bot playing the original Pokémon Red, but with a twist that the controls was by the players in the chat, with no restrictions. This meant that the bot will accept the commands of thousands of viewers’ inputs and run them all, resulting a chaotic flow throughout the entire run.

“TPP not only inspired an entire generation of Pokemon fans, but it directly inspired Twitch” – Marcus “djWHEAT” Graham, Twitch Studios director

TPP opened up the medium of streaming to new means of innovations, enabling new perspectives at interactive content. This greatly interested me back then since it memes and the cult aside, showed relations and explores giving control and goals of participants, how an exceedingly large number of participants can affect a piece of work.

Is it generative art? Generative art requires human interaction against a computer algorithm, which results in chaos and unpredictability. TPP is an extremely powerful example of thousands of human commands working against a bot that computes and returns results, often times undesirable.

How does this relate to me? My FYP leans towards chaotic player interactivity, caused by multiple players simultaneously. TPP can be said as a powerful pioneer in the direction I am headed for.