Neural Screen and intelligent patterns

The paper presents the concept of Neural Screen-an architecture constituting a working space in which a large number of intelligent patterns may navigate and interact with each other. Each of the intelligent patterns represents a piece of data to be processed as well as a strategy of its own behavior in the space. The idea of the Neural Screen may be recognized as a step towards a neural general purpose supercomputer working based on a game of intelligent patterns. An efficient strategy of pattern behavior has been obtained via ´trial and error´ method using a simulation model of an instance of the Neural Screen. The architecture may be implemented as an appropriately large set of special purpose neural chips. Since each neural chip is assumed to be connected only with its the nearest neighbors, a supercomputer might be assembled in the manner resembling playing with LEGO blocks. Owing to an assumed redundancy of navigating patterns the supercomputer should be able to continue working even in case of a partial destruction.