A model of primitive consciousness on an autonomously adaptive system

A model of primitive consciousness is proposed through investigation of a system that autonomously adapts, without a teacher, to its environment. This system is required not only to respond to the environment as fast as possible but also to provide a response of a quality that is appropriate to the situation based on its previous experiences. The system should grasp the situation, decide the action appropriately, and adapt by modifying its own configuration on the basis of its experience. To do these things as a single entity, the system has one evaluation mechanism, that is based on rewards or punishments. First, a main part of consciousness is modeled as a function to do an appropriate speedy action. For this, the system should quickly calculate a draft of a desired action based on a large amount of information, such as the results of recognition from sensory processing, recollections from the episodic memory, and states of the system itself. This calculation is done with neural loops, in which information is represented as a random pulse ratio. Each node relates to an imagery concept. Connections between nodes correspond to relations between imagery concepts. Each loop represents constraint conditions or equations. These loops operate as a solver that uses the iterative method of nonlinear simultaneous equations. The pattern of exciting nodes represents drafts of action at that time. The concept of "functional consciousness" is explained by using the above function. Second, to allow the system to control itself efficiently, only information needed to control the system is selected. The information selected, such as available alternatives or action space with relation to rewards and punishments, make up a compact space that corresponds to the everyday world that we feel consciously.