The objective of the work reported here is to create the conditions under which there can be a spontaneous increase in complexity of digital replicators. The study focuses on complexity in the sense of differentiation into different ``cell types'', where cell types are distinguished by what part of the code they execute.
The network Tierra experiment [9] attempts to create a digital environment within which replicators can evolve freely by natural selection in the digital equivalent of an ecosystem, where biotic interactions may provide a significant driving force for evolution. The original Tierra experiment [6, 7, 8] also created such an environment, however, only modest increases in complexity were observed. The network environment provides additional selective forces, which potentially could favor the kind of complexity increase that is sought in this study.
Network Tierra creates a new web on the internet. The World Wide Web is a sub-net of the internet. It is created collectively, as the result of web servers running on many machines. This collection of web servers creates an open space, within which, web browsers can move freely, accessing the data on the disks of the servers that is made available through the web server software.
The Tierra network is another web. It is created collectively as the result of running Tierra servers on many machines. The Tierra server is a piece of software written in the C language, which creates a virtual machine called Tierra. Tierra does not self-replicate, evolve, or experience mutations. Tierra does not migrate on the net. In order to run a Tierra server, someone must download the software, install it and run it.
The collection of Tierra servers creates a sub-net of the internet, within which digital organisms and Tierra browsers (Beagle) are able to move freely, accessing CPU cycles, and the block of RAM memory that is made available by the server. Note that the digital organisms and Beagle can not access other RAM on the machine, nor may they access the disk.
We can think of the web of Tierra servers as an archipelago of ``islands'' (which we usually refer to as nodes or machines on the network) which can be inhabited by digital organisms. The digital organisms are mobile, and feed on CPU cycles. Therefore, selection can potentially support the evolution of network foraging strategies.
In this experiment, we must create conditions under which selection will favor more complex migratory algorithms, over small highly optimized algorithms that only reproduce locally, such as evolved in non-network Tierra. Toward this goal we introduced the ``apocalypse'' which at random intervals kills all organisms living on a node. This provides an absolute selection against non-migratory organisms, insuring that only migratory organisms can survive in the network environment.
The tests reported here involved about 150 participating machines, located at ATR in Japan, Aizu University in Japan, The University of Delaware, The Santa Fe Institute, The Free University of Brusells, Sussex University in England, and the Swiss Federal Institute (EPFL) in Lausanne.
The Tierra process runs as a low priority background process, like a screen saver, by using a ``Nice'' value of 19. This causes the CPU cycles available to Tierra to mirror the load of non-Tierra processes on the machine (the speed of Tierra is high when the load from other processes is low). Thus the speed of Tierra will vary with the load on the machine. Also, when the user of a machine touches the keyboard, or the mouse, Tierra immediately goes to sleep for ten minutes (from the last hit). We are expecting the heterogeneity in available CPU cycles to provide selective forces which contribute to maintaining cell differentiation.