Abstract
Advances in artificial intelligence (AI) are fueling a new paradigm ofdiscoveries in natural sciences. Today, AI has started to advance naturalsciences by improving, accelerating, and enabling our understanding of naturalphenomena at a wide range of spatial and temporal scales, giving rise to a newarea of research known as AI for science (AI4Science). Being an emergingresearch paradigm, AI4Science is unique in that it is an enormous and highlyinterdisciplinary area. Thus, a unified and technical treatment of this fieldis needed yet challenging. This work aims to provide a technically thoroughaccount of a subarea of AI4Science; namely, AI for quantum, atomistic, andcontinuum systems. These areas aim at understanding the physical world from thesubatomic (wavefunctions and electron density), atomic (molecules, proteins,materials, and interactions), to macro (fluids, climate, and subsurface) scalesand form an important subarea of AI4Science. A unique advantage of focusing onthese areas is that they largely share a common set of challenges, therebyallowing a unified and foundational treatment. A key common challenge is how tocapture physics first principles, especially symmetries, in natural systems bydeep learning methods. We provide an in-depth yet intuitive account oftechniques to achieve equivariance to symmetry transformations. We also discussother common technical challenges, including explainability,out-of-distribution generalization, knowledge transfer with foundation andlarge language models, and uncertainty quantification. To facilitate learningand education, we provide categorized lists of resources that we found to beuseful. We strive to be thorough and unified and hope this initial effort maytrigger more community interests and efforts to further advance AI4Science.