A Novel Computational Artificial Intelligence Framework for Complex Physical, Chemical and Biological Networks

We have designed and implemented an efficient chemical computing architecture and platform for simulating realistic chemical systems in conventional silicon processors. The architecture uses new logic (non-boolean), computing primitives, software architecture, and a new high-level programming language. This new framework adopts the advantages of conventional computing (access to large computing power) and will be optimized for computing of complex physical chemical or biological systems.

Our network-based approach encodes reactions in the nodes and the flow of reactants to products in the edges. Our new computing model and highly modular architecture preserves the fundamental physical relationships in reactions and appears to scale more efficiently compared to currently available methods. As this appears to be the first computing model architecture tailored for chemistry and possible extension to biology, we anticipate comparing the performance of our model to current prediction-oriented methods, for which conventional machine learning models are available. The new programming and the information logic of the underlying physics can be used by the community to program different chemical systems as needed for AI systems.