Desert ants achieve reliable recruitment across noisy interactions.
Recently, physical thinking has been making progress in domains at which it did not aim originally. In this contribution, I want to sketch some examples of what can be done. The method consists of finding systems which originate in complex or complicated structure or dynamics, and which can profit from questions physicists ask. The examples I want to present comprise biology, language and the Worldwide-web (WWW). I find it fascinating that relatively simple methods, questions, and techniques from the exact sciences seem to be able to shed new light, and also new insight, into structures which are mostly self-generated. I want to suggest and illustrate that questions outside physics proper can be developed fruitfully by physicists. My story is neither totally new (see, e.g., ) nor as revolutionary as it may seem. I just want to convey my interest and pleasure in addressing "esoteric" questions with the tools of mathematics and physics. The discussion will be in the subject of "network theory," and I first summarize some of its literature [1, 12] : With the advent of powerful computers on every scientist’s desk, it has become easy to analyze large data sets. These data sets come often, and quite naturally, in the form of large networks (graphs, directed or undirected), where the nodes of the graph are certain objects, and the edges are certain binary relations between them. For example, the nodes could be individual researchers, and the links could signify that they either co-author a paper, or cite each other. Other examples are pages and links in the WWW, which connect two pages; airports and connections provided by commercial airlines; words and links between these words and their definition in a dictionary. I will call such graphs real-life graphs 1. Experimental automation, and the availability of large databases through the internet provide many interesting networks for analysis. The most useful ones are obtained in collaboration with experimental scientists. Continuing a long tradition in statistical physics, the studies of large networks often concentrate on their statistical properties. Erdős and Rényi described a set of random graphs which are built as follows : Take N nodes (N very large) and assume that the mean degree (number of links coming out of a node) is k > 0, independently of N. Then, paraphrasing Erdős and Rényi, one can make two statements: