This article was taken from the July 2014 issue of Wired magazine. Be the first to read Wired's articles in print before they're posted online, and get your hands on loads of additional content by <span class="s1">subscribing online.
There are many systems in nature, from brains to ant colonies, that are regulated without central control. Large data systems, such as the internet, work in an analogous way. Regulation without centralcontrol uses local interactions. The pattern of interactions in the aggregate produces the behaviour of the whole system.
Just as a neuron uses recent experience of electrical stimuli to decide whether to fire, an ant uses its recent experience of interactions to decide what to do. Ant colonies have one or more reproductive females, and although the ants that lay the eggs are called queens, they don't tell anyone what to do. Ants smell with their antennae, and when one ant touches another with its antennae, it can tell by the odour whether the other ant is a nestmate, and sometimes what the other ant's been doing. The important information is the pattern of interaction itself.
In different systems that are regulated without central control, there is a fit between a network's environment and how it uses interactions. Environmental constraints influence the evolution of the way that interactions regulate the system.
One important constraint is operatingcosts. Both desert ants and TCP-IP, the transmission control protocol in the internet, deal with high operating costs. Desert harvester ants have to spend water, lost when foraging in the hot Sun, to get water, which they metabolise out of seeds they collect. In the early days of the internet, operating costs were so high that it was not worthwhile to send out data if bandwidth was not available. In both systems, interaction networks are set up to generate positive feedback. The system stays inactiveunless something positive happens, and positive events increase the system's activity. A forager does not go out unless it experiences enough interactions with ants that have found food. A data packet does not go out unless returning "acks" (acknowledgements) show that previous data packets had the bandwidth to move on.
By contrast, in the tropical forest, operating costs are low for ants. One species that lives in trees sets up circuits of ants flowing constantly from nest to food-source in both directions.
Because ants are so abundant and diverse, competition is high. Many species use resources that are coveted by others. Interactions are used to generate negative feedback. The system keeps going unless something negative happens. A forager continues along the circuit unless it meets an ant of another species, in which case it is more likely to go back to the nest. An analogy with an engineered system may be a fibre-optics network that continually transmits data unless there is an interruption, or a security system that denies access only when a threshold level of incursion is reached.
We have much to learn from the ways that ants use interactions to solve problems without anyone in charge. They may provide innovative ideas for the systems that we create. They have evolved algorithms to use minimal information to regulatethe behaviour of large systems. Desert harvester ants use an algorithm similar to TCP -- and some of the other 12,000 species probably use algorithms we haven't thought of yet. There may even be interesting lessons for human interaction.
Deborah M Gordon is a professor of biology at Stanford University. Her next book, on the ecology of systems regulated without central control, will be published in 2015 by Palgrave
This article was originally published by WIRED UK
