Evolving Out of the Virtual Mud:
An Interview with Ed Burton

By Margaret Wertheim

This interview was first published in:
Cabinet, Issue 19, Spring 2005

MW: How did you come to design sodaconstructor?

EB: I had just joined Soda Creative, who were doing a lot of software in a language called Java that I’d never used before. So I needed to learn this quickly and the way I like to learn things is to play. Essentially I thought up sodaconstructor as a toy I could build in Java and play with.

MW: What kind of a toy is it?

EB: It’s in the tradition of a construction toy. You create a set of points and join them together by lines that effectively act like springs and you can play around with all sorts of configurations and see how these different models behave within the world. The world itself has some very basic physics programmed into it, such as a simple simulation of gravity. The thing I never anticipated was the diversity of forms you could build with just this simple set of rules.

MW: Can you describe how one of these models works?

EB: Sodaconstructor is a two-dimensional world populated by points and lines. The points represent masses and the lines represent springs. In the real world, springs usually have a fixed rest-length. The novel element I introduced in the software is to have some of the springs change their rest length over time. You can think of the springs as a bit like muscles expanding and contracting. The change is driven by a sine wave that is programmed into the world and acts like a clock within this space. By connecting the point masses in various configurations and letting the structure flex freely under the influence of this wave, it was possible to make simple walking creatures. I suppose that was the height of my ambition—to make simple things that walked.

MW: So you were trying to make a virtual version of a mechanical toy like Mechano or an Erector Set?

EB: Actually, it’s difficult to think of a physical toy that is like this. Most mechanical vehicles tend to have wheels because wheels are easy to build out of cogs. But typically in nature you don’t get wheels. I was thinking of something more biological.

MW: How did other users get involved with the program?

EB: At Soda Creative we originally put it up online in 1997. But it went into a very discreet corner of the site and was essentially unnoticed for several years. In the summer of 2000, we gave the site a polish and made it more prominent, then quite spontaneously it started to attract an exponentially growing amount of interest. It was largely a viral process. Suddenly it washed around the world and we started to see very quickly all these different kinds of creations coming in.

MW: Many of these models seem organic; it’s almost as if they are alive. They also seem to come in distinct types or “species.” Could you describe some of these types?

EB: One type has come to be called amoebas. Essentially, an amoeba is a ball that locomotes by squashing itself along a rotating axis. I don’t know if there’s anything in nature it would be like. That’s one of the few types that dates back to the first batch of models I made. Since then, our users have come up with many different kinds of amoebas and a lot of other types I never imagined. There’s a whole style of models based on intermeshed circles and mobius strips and there are many, many different types of walking models, far more sophisticated than anything I envisaged. For instance, bipeds. I suppose I had aspired to create bipeds, but they always struck me as much too difficult.

MW: What’s so hard about bipedal motion?

EB: When bipeds walk, they are constantly out of balance, so to achieve this style of motion requires dynamic balance with active feedback to keep from falling over. In fact, I found a forum where someone said that sodaconstructor would never create bipedal motion because there wasn’t any active feedback from the environment. But our users have managed to make very effective bipeds by exploiting the fact that as a biped falls forward, it gets a slightly better grip on the ground which enables its feet to accelerate slightly. In effect, as it falls forward, it’s able to right itself automatically. It’s interesting that users were able to make a sort of patented method for dynamic balance.

MW: Do you think they were consciously thinking of this as an engineering problem or just hit upon the method accidentally?

EB: It would have to be fairly conscious and a prolonged process of trial and error. The way things get made in sodaconstructor, it’s not that you just think, “Ah, I can make this” and draw it on the screen and there it is. Things have to be built slowly, through discovering the dynamics of individual experiments and then building up slowly to create more sophisticated things. Something like balanced bipeds requires lots of experimentation.

Another thing the users have been able to do that took a lot of conscious effort is to develop models that fly. I really didn’t expect that. They’ve made models that effectively defy the gravity built into the system. This is an example of the community going through a process of experimentation and creating something and analyzing it and improving on it. They weren’t just revealing something I’d hidden in the software; they were genuinely discovering new potential I hadn’t even suspected.

MW: It sounds quite Darwinian.

EB: It is like evolution in many ways, but it’s missing some of the mechanisms that would allow it to be really called evolution. Most importantly, it lacks any competition for resources which evolution typically has. Perhaps that’s one reason why the sorts of things you see in the soda-zoo are somewhat like what you see in the pre-Cambrian explosion in the earth’s fossil record. There’s this almost crazy diversity because there’s so little competition in the environment. Soda users think, “As long as it doesn’t fall down, it’s OK.”

MW: Looking at the evolution of the Soda universe and seeing how the models have progressed from very simple to ever more complex, do you think there is a lesson here for understanding the development of life on earth?

EB: It’s a difficult issue, because everything in the soda-zoo has been designed by humans. In that respect it’s more like creationism in that there are always conscious agents behind it. That makes it difficult to take away lessons applicable to nature, other than the idea that there is some driving force towards diversity rather than pure homogenity—it’s like an ecosystem which doesn’t work unless it has diversity. Novelty or originality is one of the main drivers and that’s maybe a lesson both in sodaconstructor and nature: diversity is good.

MW: I gather there have been discussions in your user forums about the issue of creationism versus evolution.

EB: Yes, there are a number of people playing with the software who are strongly advocating a creationist account of life. Maybe rather naively, I imagined that doing a project like this would somehow close down the debate. But people point to it as evidence of the need for there to be a conscious agent behind creation, and it was a surprise to me that the project was able to support such a diverging debate.