Neuro-Symbolic Strong-AI Robots with Closed Knowledge Assumption: Learning and Deductions

arXiv cs.AI · 6 hours ago

Researchers propose a neuro-symbolic approach for AGI robots that combines neural networks with logical knowledge representation using Belnap's 4-valued bilattice to handle unknown facts and inconsistent information. The system represents robot beliefs through axioms and logic deductions, with unknown facts at the bottom value of the truth-value ordering expanding through learning and experience over time. This framework aims to provide both human-like intelligence emulation and controlled security for robot actions through logical inference.