Printed from www.flong.com/texts/interviews/interview_ciac/
Contents © 2017 Golan Levin and Collaborators

Golan Levin and Collaborators

Interviews and Dialogues

Interview by Carlo Zanni for CIAC Magazine

CIAC Magazine is the publication of the Centre Art Contemporaine, Montreal.
Golan Levin, 16 June 2004.

Software art, generative, randomness, aleatory: often people use these words for a huge variety of digital works. Since you are an artist but also a MIT scholar, I think you are the best person to clarify these concepts for us.

Thanks. These days I'm actually located at Carnegie Mellon University, but seven years at MIT does leave its impression.

Making definitions is always a risky proposition — I'm certain that I'll offend somebody, but I'll try anyway. "Software Art" is a new form of creative practice in which the artistic medium is software code. Usually (but not always), the code itself is not the part that is exhibited to the public; instead, the artist compiles the code into an executable program, just like all of the other software you use. People obtain software art in the same ways that they obtain their ordinary business software - perhaps by a download from the Internet, or by purchasing a CD-ROM. Software art is made with a variety of intentions: Sometimes the artwork is intended to produce a beautiful, ever-changing visual result. I've done some things like this. But alot of software art has been created to challenge the concepts and conventions of software itself - some of it, like Adrian Ward's Auto-Illustrator parody of Adobe Illustrator, is even a useful tool in its own right. Finally, some very good software art has no visual result at all, or can even be a virus, like the Biennale.py virus made by the 0101.org collective. Florian Cramer and Andreas Broeckmann have written about this range of subjects quite well.

"Generative" software art, as it is usually understood today, is artwork which uses mathematical algorithms to automatically or semi-automatically generate expressions in more conventional artistic forms. For example, a generative program might produce poems, or images, or melodies, or animated visuals. Usually, the objective of such a program is to create different results each time it is executed. And generally, it is hoped that these results have aesthetic merit in their own right, and that they are distinguishable from each other, in interesting ways. Some generative art operates completely autonomously, while some generative artworks also incorporate inputs from a user, or from the environment.

"Randomness" is very hard to define, and despite my MIT degree, I'm barely qualified to do so, since the "real" definition involves some extremely advanced information-theoretic mathematics. A good working definition is that, in a sequence of random bits, each element cannot be predicted from its neighbors or from some underlying rule. Such a sequence resists compression or compact description. For example, the pattern "01010101010101010101010101010101" can be described very compactly as "16 blocks of '01' ", but the pattern "01100101001001000001010011001110", which has the same number of elements, cannot be expressed so simply, and so it is considered to have more "randomness".

"Aleatoric" artwork incorporates randomness in its execution or composition. People have used randomness as an artistic tool for a long time before computers existed! All sorts of aleatoric techniques have been used to assist in the creation of poetry, stories, music, and drawings. There have been many reasons why artists have done so. Some artists, like the Dadaists, were interested in rebelling against the strictures of academic rules, while others, like the Surrealists, or some Zen artists, wished to circumvent the censuring influence of consciousness.

I'm not sure why today's digital artists are so drawn to aleatoric uses of randomness. Probably there are a lot of different reasons; I certainly find a use for it from time to time. But I'm a little skeptical of artists who endorse an uncritical attitude towards its results. Last May, I saw a very well-known web artist give a talk at a conference called "User_Mode" at the Tate in London. He showed a piece of generative Flash art, in which everytime he clicked the mouse, he got another random arrangement of flower pictures. He was totally transfixed by his own artwork's generativity and kept clicking his mouse over and over, saying "Another one! Magnificent! Another one! Beautiful! And this one, astounding! Oh, marvellous! Miraculous! Dude, I could do this all day!" I was nauseated. His flower collages were good, but they were all equally good — and he failed to see that this made them all equally bad as well. It's one thing to endorse the beauty of unexpected outcomes, but we must confront the fact that our algorithms are capable of coldness and ugliness, too, or we will never learn anything.

Can you please tell me more about Floccugraph and the Yearbook pictures projects?

Both of these projects are aleatoric treatments of photographic source material. These projects use a photograph of a human face as a starting point, in which a person's face is treated as a probability-field for random operations. Specifically: in the "Cellular Portraits", random points are scattered across the image surface, according to a probabilistic density field governed by the darkness of the face photograph. In other words, the darker a part of the face is, the more likely that random points will be placed there. These points are then used to generate the bubble-like Voronoi diagrams you see. In "Floccugraph", the darkness of the face is used as a probability-field for generating random tugging or repulsion forces on a series of hairlike filaments. Both of these processes always turn out differently, each time they are executed.

While for JJ, and yearbook pictures you seem to be more in control (you selected the different faces for both the works), apparently Alphabet Synthesis Machine (a networked software allowing the user to generate a font-set starting from a hand drawn "seed") leaves the users more freedom to customize their font, (but they are still influenced by your algorithms). Can you please tell us something about its creation process? Can you please, in brief, unveil to us part of the generative algorithms of the software?

I think my works vary considerably in the extent to which I control their appearance, versus the extent to which their visual results are governed by autonomous algorithms, or guided by the interaction of some other user. The three projects you mention happen to lie at different extremes of this range. But it's a mistake to assume that my control, as an author, begins and ends with the extent to which I have pre-determined the precise manifestation of my works' visual results. This notion of providing or withholding control is really a sleight of hand — I may or may not have pre-determined the visual output, but I still selected and programmed the algorithms that relate the input to the output! Control has been supplanted by meta-control.

Put differently, much of the magic in creating interactive or generative art is in creating an illusion of control: the sense that the "artist" has relinquished authorship to the user, or to some clever algorithm. In fact, this is a myth. In a sense, generative artists are more in control than artists have ever been. Instead of authoring individual expressions, they now author systems for expressing. But with this power comes a greater critical burden. Traditional aesthetic criteria are not sufficient to evaluate the resulting expressions. The system itself must additionally be evaluated, according to the range and plasticity of its actual and possible expressions.

The Alphabet Synthesis Machine is an online system which allows users to "evolve" their own personal alphabet. I think it's important to mention that the resulting alphabets are nonsense-writing, which is to say, they hopefully evoke the alphabets of unfamiliar civilizations, and they probably are not readable in any customary way. These alphabets result from a process in which the user first provides a hand-drawn "seed", as you mention, and then guides the evolution of a genetic algorithm which generates the alphabetic characters. It's very similar to Darwin's idea of "Survival of the Fittest": there is a population of about 1000 alphabetic characters, each of which is a little physical simulation of a hand-drawn scribbly mark. These scribbles are continually breeding, mutating and swapping genetic material. But only the "fittest" ones survive, and it's the user who stipulates the fitness metric with my graphical interface. After enough iterations, the alphabet hopefully looks pretty unique. With respect to your questions about randomness in software art, the Alphabet Machine is a combination of a random process and a user-guided process: there is randomness in the mutations that affect the population, and in their sexual reproduction process, but at the same time, the whole system is (slowly) shaped according to the goals of the user. It's a lot like breeding a new kind of dog: it takes many generations, and some luck.

It's worth mentioning that several other people have made related projects with very different techniques; for example, Matt Chisholm's "Alphabet Soup" project uses a set of pre-rendered glyph components, like serifs and posts and hooks, that snap together to create new letterforms. His project looks more typographic, while the ASM looks more handwritten, because of the physical simulations behind the marks.

How do you personally manage randomness? I'd like to hear about your "style" using this unusual "material."

Of all of the pitfalls facing a designer of interactive systems, I think the use of computationally-generated randomness is one of the surest. The appeal of randomness, in theory, is that it can be used to introduce new “information” in order to keep an interaction fresh. Unfortunately, the problem with randomness is that it contains no information at all—and can therefore be misinterpreted as actual information by an unsuspecting human user. The user suffering from a random system may ask: “Did I just do X, or did it just happen by itself?” “Am I controlling this, or is this behavior a coincidence?” In such cases, randomness is a confusing distraction which makes it more difficult for a user to understand and master an interactive system. This can be especially problematic for systems whose fundamental appeal is predicated on a tight cybernetic coupling between human and machine. I have found that, in many cases, the use of human gesture itself eliminates the need for randomness. Human gesture, in particular, is already such a rich source of input, with its own stochastic and irregular properties, that additional randomness is hardly ever needed! Randomness can also be avoided through a deeper examination of the gesture: the use of signal analysis techniques, for example, can tease apart additional expressive dimensions latent in the user’s gesture, such as its latent periodicities, or the precise character of its frequency spectrum.

For a while, I was almost religiously committed to avoiding randomness in my artwork, and I took pride in not using it. But there was a turning point, one day when I was a student at the Media Lab. I was invited to give a demonstration of my software art to the Prime Minister of Austria. He was an enormous, towering man with a crushing handshake and a glowing, unquestionable authority. As I showed him my software, I made a point of indicating that there was no randomness in the algorithms: everything he was seeing was entirely a direct function of his own gestural movements with the mouse. He turned to me and said, "But what's wrong with randomness? Life itself is random." I thought this was a great piece of wisdom. So after all my exhortations, I somewhat embarrasingly admit to using randomness, especially in the form of low-amplitude, high-frequency statistical noise. It's basically essential for creating any kind of organic texture. Without it, most generative systems would seem lifeless, overly-regular, and dull.

These days, my opinion about randomness is in line with that of John Maeda, who wrote in Design By Numbers that "if you are going to use randomness, you should at least know where it comes from." Generative artists and designers ought to be aware that the randomness generated by a computer is NOT true randomness! "Random number generators" are actually totally predictable algorithms, and these mathematical tricks are nowhere near as perfectly chaotic as Mother Nature. The different imperfections and regularities of algorithmic random number generators, moreover, dramatically color the character of any generative artworks that use them. Unfortunately it seems that most generative artists are not aware of this.

I contend that artists who use random number generators owe it to themselves to research the properties of different random-generating algorithms. They ought to know that there are many different flavors of random number generators, such as Cauchy, Exponential, Assymetric, Gamma, Pareto, Kotz and Johnson Continuous Univariate, Discrete, Poisson, Binomial, and Linear Feedback Shift Registers. Each of these algorithms generates completely distinct kinds of randomness! I get a little upset when generative artists just rely on the standard Random function provided by Flash. It's as if there's a sale on Red paint at the art supply store, and so that's the only color people are painting with. It's one thing to use it because it's cheap and expedient, but we've gotten to the point where artists seem totally unaware that there can be other colors of randomness at all.