The fields of neuroscience and artificial intelligence have, over many
decades, respectively, established a huge collection of observed
biological facts, and developed an extensive set of theories regarding the
processing of information. However, despite this tremendous expenditure
of intellectual capital, the daunting complexity of the brain leaves
humanity at the dawn of the 21st century with an almost complete lack of
understanding of how the brain represents, stores, and processes
information. Beginning with Aristotle, philosophers and, more recently
psychologists have suggested that association forms the fundamental
cognitive processing operation. However, there is no proposed underlying
mathematical basis for investigating the brain's operation.
This talk will present the first vestiges of a theory of information
processing based on the (limited) observed operation of the thalamacortex,
and information theory. Three novel associative memory neural network
architectures will be presented, along with a set of experiments from the
field of linguistics which demonstrate a system capable of performing some
identified cognitive functions: the formation of abstract categories, the
ability to combine various sets of information of differing types, context
sensitive prediction, and the ability to determine answers consistent with
all assumed facts and all knowledge bases.