Semblions of Words — The Language of Natural and Artificial Neural Networks
Wieslaw Ludwik Galus · Qeios · 2024
The hierarchical structure of metaphors, as linguistic concepts and other mind tools embodied in Lakoff's cognitive linguistics terms, requires an indication of the mental representations in human minds and the corresponding biophysical representations. The author claims that the searched biophysical representations, which are the basis of mental representations, may be hierarchically related neurons called semblions. According to the connectionist concept, the hierarchical structure of a neural network emerges during the learning process and concurrently constitutes a memory. Semblions of objects, concepts, ideas, and models can be associated on a neurological level with neural representations of phonemes heard simultaneously while perceiving objects and creating their representations. This way, new higher-order semblions are made, corresponding to the words. The work shows how stimulating semblions can cause recall, association, thinking, and other higher mental functions necessary for using natural language. In this way, the reconstruction of mental states takes place based on the propagation of neuronal excitations. The author also indicates why semblions representing mathematical concepts adequately describe many physical world phenomena. At the same time, their polymodal counterparts can be used for a metaphorical description of qualia. The complementarity of the associative memory model proposed by Horzyk with the model of the architecture of self-conscious systems proposed by Galus, which can be used to build artificial self-conscious systems that reasonably use both natural and formal languages, was pointed out. The further development of recursive models of consciousness leading to the presentation of the Motivated Emotional Mind (MEM) model has confirmed that in the process of bottom-up information transfer responsible for creating semblions, rules of similarity are utilized from engrams representing knowledge in the mind to patterns of neural excitations transmitted from the lower sensory layers of semblions. The functioning of the multilayered architecture of brain fields based on the principles of searching for similarity patterns aligns with the hypothesis of using metaphors in subsequent processes of generalization and association, forming the basis for higher mental functions, and creating the semantics and syntax of the language.