Abstract

Current AI is trapped in the Stochastic Paradigm: high-dimensional probabilities produce hallucinations, inconsistency, and fragile reasoning. To address these issues, this paper introduces the rule-based mechanism of human logical thinking, which follows a set of universal rules to perform the corresponding types of thinking. Behind this lies a mechanism through which neural activity follows the objective interrelationships to establish the corresponding conceptual relations within the neural network. Thus, the relationships of serial, parallel, convergence, divergence and symmetry are correspondingly translated into causal thinking, parallel thinking (analogy), convergent thinking (inductive reasoning and generalization), divergent thinking (deductive reasoning) and symmetrical thinking (opposite thinking). The investigation of logical thinking is part of the much broader ontological research project on A Theory of Everything. This research reveals that the rules of logical thinking are the fundamental rules underlying everything in the universe. These rules include serial, transition, transition point, continuation, discontinuation, parallel, similarity, commonality, difference, common mechanism, convergence, divergence, contraction, expansion, symmetry, asymmetry, limitation, limitlessness, order, disorder, hierarchy and interconnectedness. These concepts can be represented by an ontological framework: a geometric model – the Fundamental Interrelationships Model (IRM) and its associated ontological-mathematical formulation. This framework can be applied to logical thinking to the various issues of AI, such as Epistemic Instability (Hallucinations), Stochastic Variance (Inconsistency), Pattern Overgeneralization, Input Fragility (Prompt Sensitivity), Recursive Decay and so on…

You can download the full, new version of the article by clicking the link: https://philpapers.org/rec/HUATMP-2