Consciousness and Subconsciousness as a Means of AGI’s and Narrow AI’s Integration

Abstract

The present article concentrates on the cognitive architecture of the agent capable to form AGI. In the process, an attempt to bridge the existing gap between AGI and narrow AI methods is made. There are two main blocks to do this - Consciousness and Subconsciousness, each of which solves the following classes of tasks: predictive analytics, prescriptive analytics, executive analytics, reflexive analytics, goal analytics, abstraction analytics, and attention analytics. Consciousness uses conscious memory and inference mechanisms. Subconscious uses models and methods of narrow AI and unconscious memory. In order to transfer information from Subconsciousness to Consciousness, emotions, insights, intuitive decisions, algorithms or sequences of actions, expectations, feelings, desires or their hierarchies, abstractions, and attention zones are used. Both conscious and unconscious memory use the metagraph model of knowledge. There is a mechanism for knowledge transfer from unconscious memory to conscious. Acquiring knowledge from the information coming from external channels is performed in parallel to conscious and unconscious memory. Similarly, information output to external channels and performing actions are based on the knowledge of both conscious and unconscious memory. AGI learning occurs due to the expansion of the number of production rules in Consciousness, as well as increasing the number of models and improving the quality of meta-learning algorithms in Subconsciousness.