Can Molecular Systems Learn?

Abstract

Research across various disciplines shows the benefits of learning and memory for gaining functionality and improving performance. It is increasingly clear that learning and memory can be found in both physical and virtual systems, from intelligent life forms to machines, simple organisms, and even designed chemical systems. We are interested in understanding to what extent physical embodiments of these processes can be synthesized and engineered from the bottom up by using molecular components. In this perspective, we raise and attempt to answer conceptual questions about supramolecular systems as the smallest units capable of learning. We define learning as a process where a complex system of interacting components modifies itself in response to an applied stress or stimulus, resulting in structural changes and information gain. We highlight the potential of systems chemistry and molecular networks to design systems that meet this definition by encoding, decoding, and storing information as memory within the system′s composition. Understanding the physical basis of molecular memory and learning could inform the development of materials and chemical systems that autonomously acquire new properties in response to their environment. This could also provide insights for next-generation computing and physical, rather than virtual, learning systems.