The Living Web – A Biologically-Inspired Multidimensional Neural Architecture for Artificial General Intelligence

Current artificial intelligence systems have achieved impressive results in specific areas, but still lack the flexible and causal reasoning abilities found in biological intelligence. This paper introduces the Living Web model, a new computational structure that combines insights from neuroscience with effective engineering methods to close this gap. The framework offers a multidimensional network topology in which nodes act as both input and output points. It uses dual-pathway causal reasoning algorithms inspired by biological dual-process theory. The architecture includes continuous learning through adjustments in synaptic strength and processing combinations of signals, creating multiple paths from different inputs. In contrast to current transformer-based and neural network models that mainly depend on statistical correlations, the Living Web model focuses on clear causal connections and the principles of embodied learning. Theoretical analysis shows that it has significant benefits over existing AI systems in terms of energy efficiency, flexibility, and reasoning skills. This positions the model as a promising step towards artificial general intelligence. It overcomes key limitations of today's AI systems while maintaining biological realism and computational practicality.