Artificial intelligence for the discovery of new functional molecules can bring enormous societal and technological progress. Here, one crucial question is how to write molecules such that computers can easily process them. In this perspective, we analyze Selfies, a relatively young method for representing molecules in a computer. Since its invention 2 years ago, Selfieshas since simplified and enabled numerous workflows for artificial intelligence (AI) in chemistry and material science.
We take an in-depth look into the future of Selfies and molecular string representations. We detail 16 new future research directions, ranging from new AI applications in chemistry, to the development of robust languages for large chemical domains, to questions about the readability of different chemical languages for humans and machines. Thereby, we hope to open a myriad of exciting doors with consequences in materials science and beyond.