Regular readers of this blog (I flatter myself that such people exist) will know I’m keen on slime moulds, a form of life that defies easy description. So the publication this week of a paper that show how a particular type of slime mould can model transportation networks in Canada was simply too good to ignore. Not only does the research explore important questions about how nature performs computations, there’s also a cool YouTube video showing a time-lapse of the cool/gross slime in action. What could be better?
The paper in question is published in the International Journal of Natural Computing Research by Andrew Adamtzky of the University of the West of England, and Selim Akl of Queen’s University. Both are in the field of “unconventional computing.” Akl describes it as follows:
We talk about building different types of computers: quantum computers, chemical computers, biological computers. Part of this involves looking at nature. For example, there are evolutionary algorithms, genetic algorithms, neural network algorithms and swarm intelligence.
Adamtzky focuses on the slime mould Physarum polycephalum. Like many slime moulds, P. polycephalum has a lifestyle that smacks of science fiction. It starts out as spores which, on exposure to dark and moist conditions, germinate into amoeba-like swarm cells. These cells reproduce asexually, and can grow flagella for swimming if they need to. When enough of them are established, they switch to sexual reproduction and form a zygote, which in turn forms forms a giant cell with many nuclei, big enough to be seen without a microscope. This is called a plasmodium, and is the source of the slime mould’s scientific name (roughly translated, Physarum polycephalum means “multi-headed slime”)
The coolest thing about the plasmodium is how it forms a network of microtubules to search for and digest food. Adamatzky has long been interested in how these networks can be seen as a computer algorithm that solves a particular problem, namely constructing an efficient set of connections between various nodes (food sources). In previous work, he’s compared P. polycephalum networks to road transportation in the UK and Mexico. But according to the latest paper written in collaboration with Akl, it’s important to model as many countries as possible, and since Akl is Canadian, the natural choice was Canada.
The setup was simple: a large petri dish was placed over a map of Canada and 16 nodes were created of top of major cities, as well as minor ones that are nonetheless major nodes in the Canadian transport system (e.g. Thompson, Man. or Wrigley, NWT). Each node consisted of a small pile of oatmeal wrapped in a paper towel, and kept moist with tap water. Would that my own graduate studies could have been conducted with such basic equipment. . .
23 times the mould was inoculated at Toronto, and after 2-5 days the network was complete. P. polycephalum never makes the same network twice, but certain connections were more common than others. For example, a connection between Thunder Bay and Winnipeg appeared in 22 of the 23 experiments, but that between Montreal and Radisson, Que. only appeared 8 times. The team applied various thresholds for the number of times a link would have to appear before it would be considered ‘strong.’ They then compared the networks thus generated to those of the actual transportation system in Canada, as well as networks generated by Akl using computer algorithms. In general, the mould imitated these very well, with few of the redundant links that one would expect to find if it was just branching out in random directions. “We would call it a computation; it’s not a brute force coverage of the whole area,” says Akl.
Just for fun, the team then simulated what would happen if the transport network was disrupted. They did this by sprinkling sea salt (toxic to P. polycephalum) at Tiverton, Ontario, site of the Bruce Nuclear Generating Station. According to the paper, the mould “migrates outside Canada, enhances the transport network outside the contaminated zone [and] sprouts indiscriminately from urban areas and transport links.” I have no doubt that the folks at Bruce Power are operating their plant in a safe manner, but in the extremely unlikely event of a nuclear meltdown, I would probably respond in much the same way.
Fun and games aside, the research into how nature computes is as important as it is fascinating. I’ll leave you with some inspirational quotes from Selim Akl himself:
Since its inception, the paradigm of computation has been: you give me the data, I’ll go and crunch the numbers, and give you the answer. There will come a time when this paradigm will change. The numbers that you are working on will change as you are working on them, and your algorithm will have to adapt to this. For example, it could be a robot roaming the surface of Mars, interacting with its environment and its environment affecting its computation.
The philosopher in me says that nature is really performing a computation: it’s computing the next state of the universe. Biologists had a description of how the world works, and then chemists came along and said it’s at the molecular level. Then physicists said it’s at the atomic level. We are saying that perhaps it’s at the bit level, or the information level. A chemical reaction is really an exchange of information between two species. If you understand it this way, maybe we can capitalize on the knowledge that we have accumulated over the past 70 or so years in computer science, which I think is quite exciting.
You can see lots more cool videos from Andrew Adamatzky’s other experiments here.
Also, in 2010 another group carried out a similar experiment with the Tokyo subway. You can see that video here.