This research will continue investigation into the theory of fault-tolerant computation. Of particular interest are devices made of a large number of elementary components each of which can fail independently of the others, with a certain constant probability. The faults are transient: they chance the local state but not the local transition rule. A fault-tolerant computer of this kind must use massive parallelism. The simplest model of a massively parallel computer is the cellular automation: a homogenous array (of dimension one, two or higher) of identical, locally communicating finite-state automata. The model is identical to the interactive particle system in statistical physics. It is an important object of investigation in statistical physics and theoretical biology as well as computer science. As the number of components that can fail increases it does not help to run two computers simultaneously (as it is done now in many critical situations), comparing their results periodically; indeed, faults will occur in both of them between comparisons with high probability. In cellular automata, the self-correction mechanism must be built into the transition rule of the cells.
