Biota.org
Mathijs Romans Interview

Mathijs Romans' Home Page

Mantrem

To give a little bit of background: I am a physicist currently pursuing a PhD in the Netherlands. I started tinkering with Alife as a student, having an interest in complex systems in general. Probably the origin of this interest is the book The quark and the jaguar by Murray Gell-Mann, that I read as a kid. I was already playing with my own Alife-programs when I read the book Artificial Life by Steven Levy, which is very interesting by itself, though it also deals a lot with cellular automata, which is in my opinion not Alife. I think it goes for a lot of people in the people that one discovers the possibilities and indeed powers of Alife by just play.

Your Zon development is stunning. It has a great retro interface and many hours of experimentation and tinkering. Zon and your previous Matrem development are listed through the GNU How-To-Alife web pages. For those not familiar with both developments, could you please give some background to them? What did you learn through developing Matrem and Zon?

Zon is really not an Alife program, because it is more like a game. It is like a real-time strategy game, where the players issue high-level commands to the inhabitants of the Zon world, that they have no control over directly. The inhabitants are quite intelligent, but do not evolve. Zon is also not finished.

The idea of Matrem is to be a playing ground for Alife. It has a sufficiently complex world with trees, grass, lakes etc. This world is inhabited by agents, creature that feed on grass and each other, and procreate. Now because of the complexity of the world, each agent needs a strategy to survive: finding new patches of grass, chasing other animals or running away, different combinations of internal parameters such as weight, strength and speed. The implementation of these strategies is done in C++, and thus leaves a lot of freedom for implementing evolving agents. Actually most of the fun me and my friends had with this game is by creating our own strategies, and creating increasingly fit agents. That is also a form of evolution, I guess.

Many in contemporary Alife development believe AI in a simulated environment is Alife. You don't believe Zon is an Alife program. What additional properties would Zon need to have to be an Alife program?

It is essential for Alife that you have independent agents that evolve by generating mutated offspring, in such way that the fittest agents survive.

How important is the aesthetic of the interface to your Alife development? In terms of user interaction and feedback, how much interface is enough for Alife?

The looks of a program by itself is not important. However, for my development I need to tinker a lot with the stuff I make, and must see the results right away. Having a long list of numbers as output makes that very difficult, so I chose some sort of graphic interface.

Okay I admit it, it is also makes it more fun.

You are developing a new Alife program currently. Can you give some background on this new program? Will it be visually similar to Matrem and Zon?

Yes, I have little time now, but I do have some ideas. After Matrem I made a program that simulated the development of trees, which resulted in quite realistic forests. I claim success because I could see a very clear forming of niches, i.e. forking of species of plants.

Right now I am working on a whole different concept of evolution: economic and social evolution. It has been noted many times that the economic struggle of individuals and companies is also very much like an evolutionary system. So at the moment I develop a simulated environment where people live, work and trade. There is a currency system, and freely evolving prices for goods. Well managed corporations will grown and prosper, others will fail and close down. Eventually I could evolve social structures by introducing theft, feudalism etc. It will be visually very different to Matrem and Zon, as the interface is now 3D.

Your sites have Go related links and software too. There is a good subset of Alife developers who also have an interest in Go. What do you think is the correlation between Alife developers and Go?

I really think there is a connection: Alife studies a method to solve an abstract and complex problem in a systematic way. Go is also very abstract and fuzzy, and is very difficult to tackle in a logical way. I also think that people from the west that decide to study Go tend to be people that like to think outside the box.

As you are completely self taught, have you had any interaction with the Alife community? What is your sense of the broader Alife community?

Well, trough the internet of course. I learn from all the nice projects I see that are going on. It's quite active, but I am not sure about concerted scientific efforts to propel our knowledge. I have a feeling that this may be lacking in the field, but I am really not the one to know.

What more would you like to see with the Alife community?

I would like to see some concrete examples of difficult, abstract problems that are relevant to our society, and can be solved by Alife. Then it could become much more an alternative to AI, in stead of being just a part of it.

Any final thoughts?

Yeah, check out Matrem at http://www.phys.uu.nl/~romans/matrem/ or in Debian/Ubuntu repositories!

Many thanks for the opportunity to talk with you.

The interview was taken by Biota.org's Tom Barbalet via email on March 25th, 2006.