We are building systems at the human level and beyond, drawing from humanity's deepest insights into cognition itself. Our OpenCog Hyperon framework synthesizes knowledge across diverse fields, including philosophy of mind, cognitive science, computer science, mathematics, and linguistics, to create truly human-like artificial intelligence.

This requires a cognitive modeling approach that focuses on functional, rather than simply structural, similarities with human cognition. We have achieved this in our Economic Attention Networks (ECAN) module for attention and resource allocation, originally implemented and tested in the OpenCog AGI framework. Within ECAN, working memory, as implemented in the PRIMUS cognitive architecture, is stored as Atoms that possess sufficiently high Short-Term Importance (STI) values.

We are currently porting ECAN to OpenCog’s newer, more streamlined, efficient, and scalable Hyperon successor. Concurrently, we are exploring a variety of novel mechanisms incorporating neural predictive coding, generalizations of Bennett’s set-theoretic ”weakness” concept, and complex dynamical systems, among others.