A multi-decade theoretical exploration of artificial and natural general intelligence is reviewed, covering patternist philosophy of mind, formal definitions of intelligence, and a proposed high-level architecture for AGI systems. The review details how cognitive processes like logical reasoning, program learning, clustering, and attention allocation can be implemented within this architecture, emphasizing a common knowledge representation (typed metagraph) to enable cognitive synergy between processes. Human-like cognitive architecture is presented as a manifestation of these general principles, with discussions of machine consciousness and machine ethics. Practical lessons for implementing advanced AGI in frameworks like OpenCog Hyperon are briefly considered.
A new conceptual foundation for information is proposed, based on "distinction graphs" where two nodes are connected if an observer cannot distinguish them. Graphtropy, defined as the average connection probability, equals logical entropy when the graph consists of fully connected components. The framework extends to probabilistic and quantum versions, linking graphtropy to thermodynamic and quantum entropy. Complex intelligence corresponds to intermediate graphtropy states, associated with memory imperfections that violate assumptions of the Second Law of Thermodynamics. For computable nodes, graphtropy relates monotonically to average algorithmic information. A quantum version yields graphtropy as a measure of mixed-state impurity. Dynamic Distinction Graphs add causal links to model observers.