Open Journal Montreal

I met this funny, soft-spoken gentleman at a sushi party being thrown by his roomates and found out, in my nosy way, that he’s a PhD in robotics. He sweetly agreed to answer some questions for me, so I emailed him a small stack of queries to get started. Here they are, and here are his answers. All of this material connects in crazy, mathy-y, robot ways with the ideas about open source intelligence and cooperation, and communication distortion and bias, we explore here at Open and, in a more focused way, in my MA thesis. So that’s cool. I’ll defintely have more questions for him, based on this glimpse he’s given us into his ideas, so stay tuned…

R: So, how did you find your way into robotics?

J: After a degree in maths, I got interested in the concept of swarm intelligence, which is a way to describe how a collection of individuals is able to accomplish more than the mere sum of their limited abilities. The phenomena at work is commonly called “emergence”, in the sense that the collective capabilities “emerge” from the interactions of individuals.

Then, because of their strong limitations in means of perception and action, I chose to study this kind of phenomena using robots. Considering that for an autonomous robot even getting its position is difficult, there is not much space left to “cheat” on the so-called amazing results you can always produce in
simulation by assuming that some crucial information is available. In working with robots you really put a strong constraint on yourself.

R: Could you talk about what an algorithm is, and could you use metaphors? if, for example, someone doesn’t much think about math but relates to the world largely through language, or music, or agriculture, how would you explain what you specialize in to them?

J: The common metaphor that is used when talking about an algorithm is the cooking recipe: a list of actions to take one after the other and possibly repeat -depending on whether the muffins are cooked or not- to get the final result, be it edible or only a sequence of numbers.

What I am more interested in is…

… the interaction of a collection of these recipes going on simultaneously. You could imagine a huge kitchen with a lot of cooks all going their way according to what they think should be done now, without any chef to say what to do or in which order. The idea is that there is a recipe for
each single cook to follow that together with all the others will bring a complex and delicate dish.

Why should we actually care about this idea considering that a delicate and complex dish can be made through the supervision of a proper Chef ? The reason is that the Chef is only going to be able to coordinate a limited amount of assistants. What if the task requires many more people than the coordination abilities of the Chef ? Human societies mainly answer this question by creating complex hierarchies. Insects such as termites, ants or bees rely on individual autonomy with absolutely amazing results.

If the cooking metaphor doesn’t work with you, the thing I am interested in, translated into music discourse, would be investigating the question of how do you get a really amazing piece of music when a million instruments each play a definite score without a conductor ? Maybe the more related idea is thinking about
language: languages can be seen as equilibriums resulting from the constant and uncoordinated interaction of individuals following grammar rules (and sometimes not following, as the ants that don’t always are fully coordinated…).

R: what’s the idea behind the project you’re in now?

The project I am in now is interested in studying the implications of the swarm intelligence in building a structure. Termites are well known for the mounds they build, which are architecturally very complex and efficient in terms of cooling, for instance.

In the project we use flying cubical blimps to investigate autonomous self-assembly of structures. In other words, the blimps are going to glue themselves together to form larger structures. However, none of the blimps has a plan of the resulting shape. This shape “emerges” from the interactions between the blimps reacting to local assembly configurations.

Many people have investigated such self-assembly algorithms with variable success, but no one has ever tried that with flying robots. So in this sense it is quite challenging. The concept of undirected yet useful architecture is really interesting for me in these times of centralised town planning resulting in poor quality of life, for instance. But the implications are much larger.

R: What do you think about collaborations between artists and scientists? What kind of recommendations would you make to people interested in embarking on a cross-discipline collaboration like yours?

J: I personnally think there is not so much a difference between an artist and a scientist. Each one needs creative abilities to make progress and present new work. I think problems arise more because of misconceptions and prejudices.

My advice before embarking would be to try to understand what are the needs and the interests of one part in getting the other involved. Good collaborations don’t come up like that but “emerge” from one part getting really interested into the other’s work. With the growing hype on science-art projects, I see as a
(pseudo-)scientist a lot of artistic work with either a “technology” or “science” stamp on it that would rather gain in concentrating a bit more on the artistic side. But I am only a scientist.

R: You mentioned that you build on linux, can you explain why? How does open source software interact with the field of robotics?

When you work on the controller of a robot, you want to be able to know why it crashed… More seriously, although not originally designed for it, unix has strong potential for real time control. The open source version linux has the advantage to be downsizeable to fit in small and limited controllers without changing much. It is called “embedded linux”. And the number of users finding out bugs and correcting them makes it the more reliable system anyway. And it’s free and everybody uses it and… The list of advantages is endless.

R: You said open source code (or maybe you were talking specifically about linux) had an aesthetic, could you explain how?

In my opinion you can see open source code as an undertaking reminiscent of medieval cathedrals. In medieval times, people would look all across Europe for the best artist or craftman to carve this very part of the transept. This resulted in a collective piece of art and craft. Written computer code can be done in an
infinite number of ways and when you start looking deep in the numerous libraries of a distribution you realise how every line is the work of an individual with its own style. And a complete distribution is really huge. It is the collection of a lot of small and big improvements going on and on, with the only purpose of making it better. It has this kind of beauty.
to read a classic bit of open source theory that uses a similar metaphor in a different way, check out Eric Raymond’s The Cathedral and the Bazaar

R: What kind of project would you want to attempt with unlimited funding?

I would like to study morphogenetic systems in the context of architecture and design.

Interviewee: Julien Nembrini, interviewer: Risa D.

tags:   

[ originally posted on Thursday, September 15th, 2005 at 4:31 pm and is filed under Communication, Everything, Open Source. ]