An Artificial Brain that Can Learn and Use Any Language

Abstract

Based on findings in quantitative linguistics for natural network structures a neural network system has been developed recently that can learn and process any coherent language text. It is essentially a technical network depending on a new understanding of information. By using measured text structures it is also a possible model for the unknown organization of neurons for thinking in the brain. In the following, the construction of the language system is outlined, based on natural network structures for word successions that were published some years ago in Glottometrics. Codeless representations of words and sentences through neurons are introduced. After that a hierarchical network system is developed for the abstraction of coherent text from level to level in the system. This leads eventually to a highly compressed representation of one or more sentences through single neurons. These new operations enable the development of novel electronic machines as e.g. perfect machine translators, etc. Last but not least the findings may also point to the possibility of similar operations in the human mind. (For example, up to now the search for codes in human neurons was in vain and therefore there is no idea, how neural networks in the brain are able to produce thoughts)