Visual Trends in AI Art
From Snapped Spiderman to Data Viz: How Particle Systems Came to Dominate Graphic Media
Whenever something smokes, smolders, flows, or flocks, it’s a safe bet there’s a particle system at work
What does data look like? For many people, the word conjures images of spreadsheets, JSON files, or reams of paper printed out on continuous stationery. But in recent years another visual language for data has emerged, eclipsing all others: dots.
When you see a screen or a page with loads and loads of dots or other identical little particles, you’re about to be hit with some big-time Data. The aesthetic of dots has become the go-to vernacular for designers and artists working with data or interactive media (in which live interactions are really just a way of capturing data in real-time.)
Dot-language has become so prevalent that there is now a growing backlash to it among creators and spectators alike. It’s difficult to parse — looking at many projects in this aesthetic, you can’t tell if the data in question is about butterfly migrations or drone strikes. And once you have determined the topic of the piece in question, you still have to make sense of the data. What do all these little dots mean? What manner of divination is required to decipher this scatter of information? And most importantly, why do so many data viz projects look Like That?
The answer to this last question has much to do with the technology that is used to produce them: particle systems.
You may not be familiar with the term “particle system,” but you’ve almost certainly seen them in action. Particle systems feature heavily in the world of CGI, and can be found in nearly every sci-fi, action, or horror movie (and many other genres as well!), the most famous example of which is probably the “Thanos snap” moment from Avengers: Infinity War, when half of our favorite heros seem to disintegrate into nothingness. Each mote of dust in these shots is a particle, and the logic governing their animation is a system — hence: “particle system.”
The CGI community uses particle systems with much affection, and not just to render explosions or scatter Spiderman to the wind. Whenever something smokes, smolders, flows, or flocks, it’s a safe bet there’s a particle system at work. One particle system called Perlin noise (named for its creator Ken Perlin) even won an Academy Award for Technical Achievement in 1997, in recognition of its ability to generate natural-looking organic textures.
Perlin developed his algorithm in the early 1980s, but for many years Perlin noise and other algorithms like it were too computationally intensive for most artists to use without the support of a well-funded institution or studio. But when consumer-grade computers became powerful enough to perform the multi-dimensional calculations that particle systems require, we saw these systems proliferate throughout media culture. At the height of the data art movement between 2011–2018 the Javascript library d3.js became a household name for artists and designers, and was the technological darling of the publishing industry from the NYT to the Guardian, and beyond.
A major figure in developing this new visual vocabulary is Blacki Migliozzi. As graphics editor for the New York Times, Blacki really turned all of us onto hip ways of representing the news into data points. While he didn’t make particle system art per se, he did take the art of data visualization into the 21st century. His work has been instrumental in establishing many conventions that have become commonplace, such as the practice of representing data points as individual particles which make up a larger whole. This visual framework laid the stage for particle systems to become intelligible representations of large data sets and allowed us to derive meaning from the ways in which those dots moved or didn’t move.
Particle systems also found use in representing concepts from the burgeoning field of machine learning. They enable us to see t-SNE clusters from unsupervised learning models. With particle systems we can use our human vision to actually interrogate data in new ways, seeing and watching correlations that the neural nets are able to surface.
The interactive documentary CLOUDS offers “an endless, ever-changing conversation… presented through real-time interactive visual systems.” The producers of this film used point clouds — a type of particle system — to depict human subjects volumetrically moving through space. This use of particle systems not only brings a visual continuity between its subjects and their work, but also maintains the privacy of incarcerated subjects, or those who wished to remain anonymous for other reasons.
At the other end of the spectrum are projects like Luke DuBois’s State of the Union, which challenges the clean, impersonal nature of the dot representation. When these techniques are applied to topics such as violent crimes, the format can feel overly sanitized and dehumanizing, as it turns victims of gun violence, for instance, into dots and pixels, de-emphasizing the complexity and dignity of their lives. Instead, Luke’s work has used guns firing every time there is an act of gun violence, or employed sound to press the images of the scatter plots of graphs of gun violence into scatters that look more like blood splatters. Although DuBois has used digital particle systems in his work, this analog version of a particle system challenges our assumptions about data representation and can carry with it a new meaning.
Perhaps the most famous particle system artist is Refik Anadol. His work has particles drawing cities, representing oceans, forming rich abstract textures, and moving in ways that have only recently become possible thanks to new physics engines in 3D.
Nancy Baker Carhill is another artist who is pushing particle systems in interesting directions. Her piece “Mushroom Cloud” has spores of mushrooms that represent a communication network turning a mushroom cloud into a regenerative mycelium network in the sky, exploring how particle systems operate in nature.
All of these artists use particles to their own imaginative ends, yet there remains a marked visual similarity across their work thanks to the technical principles involved. This isn’t necessarily a problem, although it does point to a potential limitation of the form. Particle systems are very well-suited to visualizing structures that have their own inherent logic; they can show emergent trends in large empirical datasets, and display the hidden shapes of algorithms.
Every particle in the system is identical to the next, blank little slates ready to receive attributes from some external source. This is what makes them so powerful for telling stories about emergence, but it’s also what makes them ill-suited to tell stories whose characters are qualitatively different from one another.
This is part of a 4 part series on visual trends and aesthetics of AI Art, if you liked this one, check out the first article in the series Visual Trends in AI Art.
