A Thicket of Questions: On Matthew Battles’s Earth Measurer

In the interest of discovering what goes on at Harvard’s metaLAB, I sat down for a talk with Associate Director Matthew Battles, who is working on so many projects he referred to them as “projects that nest in projects.” One project in particular he wanted to discuss is Earth Measurer, as it encapsulates his varied interests in technology, art, and the natural world.

The idea for the project was sparked when he noticed how computer scientists refer to algorithms as “a mystery” like human consciousness, Battles told me. We don’t really understand how it happens, they say. Around the same time, Battles was looking at the proliferating scientific names of organisms, which biologists ordered into taxonomic groups. He’d been focusing in particular on the group of moths whose caterpillars are known as inch worms — in taxonomic terms, the Geometridae, a name which roughly translates as “earth measurer.” Inch by inch, they measure the world, and so Battles saw a connection to how algorithms are supposed to measure, analyze, and repair the world.

From there, he had the idea to feed lists of genus names of the geometrid moths into an open-source machine learning algorithm (created by Tesla’s Director of Artificial Intelligence, Andrej Karpathy), in order to spit out similar names. When the algorithm produces a name that it “thinks” sounds like an accurate genus name, it builds on it, thousands and thousands of times, in a selective system called a neural network — a type of machine-learning algorithm patterned after the neural pathways of our brain. Through the action of the algorithm, two thousand genus names springboard the generation of hundreds of thousands of new ones. For Battles, these names represent the immensity of biodiversity. “You can shake a tree in the Amazon,” he told me, “and most of the species that fall off will be unknown to scientists.” But they also bring to mind the daily loss of such biodiversity, the unglimpsed possibility that exists in the world.

While Battles was telling me about Earth Measurer, I started to understand how a metaLAB project starts out playfully, driven by a kind of hacker’s curiosity that maybe isn’t looking for a certain answer, or even asking a specific question. Then again, there is a specific yet really basic question Battles asks, and that is: “Can I get a machine-learning algorithm to do some work for me? I mean really basic stuff, like, can I figure out how to install and run this kind of esoteric algorithm on my own computer, then tweak it a little bit here and there to do something different?” Once he’s done that, he sees what he can make out of the stuff that program outputs. But it starts out as tinkering, improv, and experiment — play. Later, the products of that playful exploration can range across all kinds of questions, like those asked by scholars, technologists, designers, and artists.

Now that the project is off the ground and beyond the initial experimental stage, I asked Battles what’s the question at the heart of Earth Measurer. He talked about the era of climate change, and how we’re aware of things like habitat destruction, extinction, and biodiversity loss — but only dimly aware of them, “as abstract issues,” he said, “not visceral matters of concern.” At the same time, the very technologies that contribute to humankind’s destructiveness — and we have to count computers among those technologies, he said — are becoming more powerful, working ever more independently from our direct oversight and control. “So I guess the question looks something like this: how does nonliving artificial-intelligence technology relate to the precarious state of life on earth? What happens when we put them into dialogue with one another?”

From researching this question, Battles has given presentations and made mini-exhibits — one of which features origami butterflies, which you open to find a bot-written message inside. Clearly, Battles is an artist at heart, and he said he was using “art tricks” to make “layers of gesture” to raise questions about what’s going on in the world. He’s using algorithms and computer learning for neither analysis nor prediction, but for poetry. In doing so, he wants to “estrange some of the normative habits of mind” we have around how we perceive the natural world, and he hopes that his work “coughs up some new forms of attention.”

One presentation Battles gave is at a Digital Humanities (DH) conference. When metaLAB began in 2011, is was conceived of as a DH lab. DH was gaining momentum then, with many in the social sciences and arts forming an interdisciplinary movement — a discipline in itself, that is — that used digital tools to do scholarship. But since then, as DH has become more focused on analysis, with many scholars using their findings as a means to get tenure, Battles said metaLAB has gone in the other direction, departing from DH by using digital tools to serve artistic practice more than scholarly analysis. And here I had an ah-ha! moment. Because one thing metaLAB uniquely does, as Battles phrased it, is “use technology to make art about science.”

That indeed might be a metaLAB thing, but I wondered if this combination was deliberate, because he was making a project for metaLAB, or simply an expression of his personal mode of inquiry and research. That is, I asked how approaching a question as a metaLAB project changes the way he works the question. “It’s an interesting question,” Battles said, “because it’s about the role and nature of questions. Now, by nature, I’m a question fan, an inquirophile, an asker.” He said he often thinks of a short verse by the French poet René Char that brings him up short, however: “What bird has the heart to sing in a thicket of questions?”

Battles wears many hats, and so it’s easy to come up with a thicket of questions. As a traditional essayist and writer of scholarly articles, he might pursue the question at the heart of Earth Measurer by doing literary and historical survey of the meanings of biodiversity and technology, and how they have been compared and contrasted through time. For instance, how did early computer scientists use living organisms as models for technical systems? That kind of question is a valid starting point for an academic inquiry. But then, as a designer and technologist, he said he might frame a different kind of question: “How might we use machine learning to analyze or predict, say, habitat loss and extinction? Or: what data visualization methods might convey the problem intuitively and concretely for different audiences?” And as an artist, the kinds of questions he might ask look very different. Like what? I asked. “Like: how is an algorithm like a butterfly?” Battles said, and went on: “What would a world of butterly algorithms look like, feel like? Can we imagine algorithms taking the place of butterflies?”

Whoa. That question was like a wrench in the gears of my brain. Because it forced me to look at things I thought I knew in a totally fresh way, from scratch. So of course it’s difficult to understand, much less explain, what metaLAB does, because not only are they taking seriously all kinds of questions, as well as the modes behind them, they’re then finding completely original ways to put those questions in conversation with each other. “The only danger,” Battles said, “is that, pretty soon, you find yourself in a thicket!”

—Randy Rosenthal