Science-Based AI, or SBAI, is a novel method of creating ML networks that understand physical systems including materials, devices, and processes well enough to be able to predict their behavior and optimize their performance.  

Just as a scientist or engineer does, SBAI incorporates all relevant and available information. In addition to the data, empirical or simulated, that a conventional ML uses, Science-Based AI uses relevant scientific laws, principles, system specifications and other constraints to better model the product being developed.