Computing with 10,000-bit words

Today's computers compute with numbers and memory pointers that hardly ever exceed 64 bits. With nan-otechnology we will soon be able to build computers with 10,000-bit words. How would such a computer work and what would it be good for? The paper describes a 10,000bit architecture that resembles von Neumann's. It has a random-access memory (RAM) for 10,000-bit words, and an arithmetic-logic unit (ALU) for “adding” and “multiplying” 10,000-bit words, in abstract algebra sense. Sets, sequences, lists, and other data structures are encoded “holographically” from their elements using addition and multiplication, and they end up as vectors of the same 10,000-dimensional space, which makes recursive composition possible. The theory of computing with high-dimensional vectors (e.g. with 10,000-bit words) has grown out of attempts to understand the brain's powers of perception and learning in computing terms and is based on the geometry and algebra of high-dimensional spaces, dynamical systems, and the statistical law of large numbers. The architecture is suited for statistical learning from data and is used in cognitive modeling and natural-language processing where it is referred to by names such as Holographic Reduce Representation, Vector Symbolic Architecture, Random Indexing, Semantic Indexing, Semantic Pointer Architecture, and Hyperdimensional Computing.